The GNU Awk User's Guide

Next: , Up: (dir)

General Introduction

This file documents awk, a program that you can use to select particular records in a file and perform operations upon them.

Copyright © 1989, 1991, 1992, 1993, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2007 Free Software Foundation, Inc.

This is Edition 3 of GAWK: Effective AWK Programming: A User's Guide for GNU Awk, for the 3.1.6 (or later) version of the GNU implementation of AWK.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with the Invariant Sections being “GNU General Public License”, the Front-Cover texts being (a) (see below), and with the Back-Cover Texts being (b) (see below). A copy of the license is included in the section entitled “GNU Free Documentation License”.

  1. “A GNU Manual”
  2. “You have freedom to copy and modify this GNU Manual, like GNU software. Copies published by the Free Software Foundation raise funds for GNU development.”

Short Contents

Table of Contents

Next: , Previous: Top, Up: Top


Arnold Robbins and I are good friends. We were introduced 11 years ago by circumstances—and our favorite programming language, AWK. The circumstances started a couple of years earlier. I was working at a new job and noticed an unplugged Unix computer sitting in the corner. No one knew how to use it, and neither did I. However, a couple of days later it was running, and I was root and the one-and-only user. That day, I began the transition from statistician to Unix programmer.

On one of many trips to the library or bookstore in search of books on Unix, I found the gray AWK book, a.k.a. Aho, Kernighan and Weinberger, The AWK Programming Language, Addison-Wesley, 1988. AWK's simple programming paradigm—find a pattern in the input and then perform an action—often reduced complex or tedious data manipulations to few lines of code. I was excited to try my hand at programming in AWK.

Alas, the awk on my computer was a limited version of the language described in the AWK book. I discovered that my computer had “old awk” and the AWK book described “new awk.” I learned that this was typical; the old version refused to step aside or relinquish its name. If a system had a new awk, it was invariably called nawk, and few systems had it. The best way to get a new awk was to ftp the source code for gawk from gawk was a version of new awk written by David Trueman and Arnold, and available under the GNU General Public License.

(Incidentally, it's no longer difficult to find a new awk. gawk ships with Linux, and you can download binaries or source code for almost any system; my wife uses gawk on her VMS box.)

My Unix system started out unplugged from the wall; it certainly was not plugged into a network. So, oblivious to the existence of gawk and the Unix community in general, and desiring a new awk, I wrote my own, called mawk. Before I was finished I knew about gawk, but it was too late to stop, so I eventually posted to a comp.sources newsgroup.

A few days after my posting, I got a friendly email from Arnold introducing himself. He suggested we share design and algorithms and attached a draft of the POSIX standard so that I could update mawk to support language extensions added after publication of the AWK book.

Frankly, if our roles had been reversed, I would not have been so open and we probably would have never met. I'm glad we did meet. He is an AWK expert's AWK expert and a genuinely nice person. Arnold contributes significant amounts of his expertise and time to the Free Software Foundation.

This book is the gawk reference manual, but at its core it is a book about AWK programming that will appeal to a wide audience. It is a definitive reference to the AWK language as defined by the 1987 Bell Labs release and codified in the 1992 POSIX Utilities standard.

On the other hand, the novice AWK programmer can study a wealth of practical programs that emphasize the power of AWK's basic idioms: data driven control-flow, pattern matching with regular expressions, and associative arrays. Those looking for something new can try out gawk's interface to network protocols via special /inet files.

The programs in this book make clear that an AWK program is typically much smaller and faster to develop than a counterpart written in C. Consequently, there is often a payoff to prototype an algorithm or design in AWK to get it running quickly and expose problems early. Often, the interpreted performance is adequate and the AWK prototype becomes the product.

The new pgawk (profiling gawk), produces program execution counts. I recently experimented with an algorithm that for n lines of input, exhibited ~ C n^2 performance, while theory predicted ~ C n log n behavior. A few minutes poring over the awkprof.out profile pinpointed the problem to a single line of code. pgawk is a welcome addition to my programmer's toolbox.

Arnold has distilled over a decade of experience writing and using AWK programs, and developing gawk, into this book. If you use AWK or want to learn how, then read this book.

     Michael Brennan
     Author of mawk

Next: , Previous: Foreword, Up: Top


Several kinds of tasks occur repeatedly when working with text files. You might want to extract certain lines and discard the rest. Or you may need to make changes wherever certain patterns appear, but leave the rest of the file alone. Writing single-use programs for these tasks in languages such as C, C++, or Pascal is time-consuming and inconvenient. Such jobs are often easier with awk. The awk utility interprets a special-purpose programming language that makes it easy to handle simple data-reformatting jobs.

The GNU implementation of awk is called gawk; it is fully compatible with the System V Release 4 version of awk. gawk is also compatible with the POSIX specification of the awk language. This means that all properly written awk programs should work with gawk. Thus, we usually don't distinguish between gawk and other awk implementations.

Using awk allows you to:

In addition, gawk provides facilities that make it easy to:

This Web page teaches you about the awk language and how you can use it effectively. You should already be familiar with basic system commands, such as cat and ls,1 as well as basic shell facilities, such as input/output (I/O) redirection and pipes.

Implementations of the awk language are available for many different computing environments. This Web page, while describing the awk language in general, also describes the particular implementation of awk called gawk (which stands for “GNU awk”). gawk runs on a broad range of Unix systems, ranging from 80386 PC-based computers up through large-scale systems, such as Crays. gawk has also been ported to Mac OS X, MS-DOS, Microsoft Windows (all versions) and OS/2 PCs, Atari and Amiga microcomputers, BeOS, Tandem D20, and VMS.

Next: , Up: Preface

History of awk and gawk

Recipe For A Programming Language

1 part egrep 1 part snobol
2 parts ed 3 parts C

Blend all parts well using lex and yacc. Document minimally and release.

After eight years, add another part egrep and two more parts C. Document very well and release.

The name awk comes from the initials of its designers: Alfred V. Aho, Peter J. Weinberger and Brian W. Kernighan. The original version of awk was written in 1977 at AT&T Bell Laboratories. In 1985, a new version made the programming language more powerful, introducing user-defined functions, multiple input streams, and computed regular expressions. This new version became widely available with Unix System V Release 3.1 (SVR3.1). The version in SVR4 added some new features and cleaned up the behavior in some of the “dark corners” of the language. The specification for awk in the POSIX Command Language and Utilities standard further clarified the language. Both the gawk designers and the original Bell Laboratories awk designers provided feedback for the POSIX specification.

Paul Rubin wrote the GNU implementation, gawk, in 1986. Jay Fenlason completed it, with advice from Richard Stallman. John Woods contributed parts of the code as well. In 1988 and 1989, David Trueman, with help from me, thoroughly reworked gawk for compatibility with the newer awk. Circa 1995, I became the primary maintainer. Current development focuses on bug fixes, performance improvements, standards compliance, and occasionally, new features.

In May of 1997, Jürgen Kahrs felt the need for network access from awk, and with a little help from me, set about adding features to do this for gawk. At that time, he also wrote the bulk of TCP/IP Internetworking with gawk (a separate document, available as part of the gawk distribution). His code finally became part of the main gawk distribution with gawk version 3.1.

See Contributors, for a complete list of those who made important contributions to gawk.

Next: , Previous: History, Up: Preface

A Rose by Any Other Name

The awk language has evolved over the years. Full details are provided in Language History. The language described in this Web page is often referred to as “new awk” (nawk).

Because of this, many systems have multiple versions of awk. Some systems have an awk utility that implements the original version of the awk language and a nawk utility for the new version. Others have an oawk version for the “old awk” language and plain awk for the new one. Still others only have one version, which is usually the new one.2

All in all, this makes it difficult for you to know which version of awk you should run when writing your programs. The best advice I can give here is to check your local documentation. Look for awk, oawk, and nawk, as well as for gawk. It is likely that you already have some version of new awk on your system, which is what you should use when running your programs. (Of course, if you're reading this Web page, chances are good that you have gawk!)

Throughout this Web page, whenever we refer to a language feature that should be available in any complete implementation of POSIX awk, we simply use the term awk. When referring to a feature that is specific to the GNU implementation, we use the term gawk.

Next: , Previous: Names, Up: Preface

Using This Book

The term awk refers to a particular program as well as to the language you use to tell this program what to do. When we need to be careful, we call the language “the awk language,” and the program “the awk utility.” This Web page explains both the awk language and how to run the awk utility. The term awk program refers to a program written by you in the awk programming language.

Primarily, this Web page explains the features of awk, as defined in the POSIX standard. It does so in the context of the gawk implementation. While doing so, it also attempts to describe important differences between gawk and other awk implementations.3 Finally, any gawk features that are not in the POSIX standard for awk are noted.

This Web page has the difficult task of being both a tutorial and a reference. If you are a novice, feel free to skip over details that seem too complex. You should also ignore the many cross-references; they are for the expert user and for the online Info version of the document.

There are subsections labelled as Advanced Notes scattered throughout the Web page. They add a more complete explanation of points that are relevant, but not likely to be of interest on first reading. All appear in the index, under the heading “advanced features.”

Most of the time, the examples use complete awk programs. In some of the more advanced sections, only the part of the awk program that illustrates the concept currently being described is shown.

While this Web page is aimed principally at people who have not been exposed to awk, there is a lot of information here that even the awk expert should find useful. In particular, the description of POSIX awk and the example programs in Library Functions, and in Sample Programs, should be of interest.

Getting Started, provides the essentials you need to know to begin using awk.

Regexp, introduces regular expressions in general, and in particular the flavors supported by POSIX awk and gawk.

Reading Files, describes how awk reads your data. It introduces the concepts of records and fields, as well as the getline command. I/O redirection is first described here.

Printing, describes how awk programs can produce output with print and printf.

Expressions, describes expressions, which are the basic building blocks for getting most things done in a program.

Patterns and Actions, describes how to write patterns for matching records, actions for doing something when a record is matched, and the built-in variables awk and gawk use.

Arrays, covers awk's one-and-only data structure: associative arrays. Deleting array elements and whole arrays is also described, as well as sorting arrays in gawk.

Functions, describes the built-in functions awk and gawk provide, as well as how to define your own functions.

Internationalization, describes special features in gawk for translating program messages into different languages at runtime.

Advanced Features, describes a number of gawk-specific advanced features. Of particular note are the abilities to have two-way communications with another process, perform TCP/IP networking, and profile your awk programs.

Invoking Gawk, describes how to run gawk, the meaning of its command-line options, and how it finds awk program source files.

Library Functions, and Sample Programs, provide many sample awk programs. Reading them allows you to see awk solving real problems.

Language History, describes how the awk language has evolved since first release to present. It also describes how gawk has acquired features over time.

Installation, describes how to get gawk, how to compile it under Unix, and how to compile and use it on different non-Unix systems. It also describes how to report bugs in gawk and where to get three other freely available implementations of awk.

Notes, describes how to disable gawk's extensions, as well as how to contribute new code to gawk, how to write extension libraries, and some possible future directions for gawk development.

Basic Concepts, provides some very cursory background material for those who are completely unfamiliar with computer programming. Also centralized there is a discussion of some of the issues surrounding floating-point numbers.

The Glossary, defines most, if not all, the significant terms used throughout the book. If you find terms that you aren't familiar with, try looking them up here.

Copying, and GNU Free Documentation License, present the licenses that cover the gawk source code and this Web page, respectively.

Next: , Previous: This Manual, Up: Preface

Typographical Conventions

This Web page is written using Texinfo, the GNU documentation formatting language. A single Texinfo source file is used to produce both the printed and online versions of the documentation. Because of this, the typographical conventions are slightly different than in other books you may have read.

Examples you would type at the command-line are preceded by the common shell primary and secondary prompts, ‘$’ and ‘>’. Output from the command is preceded by the glyph “-|”. This typically represents the command's standard output. Error messages, and other output on the command's standard error, are preceded by the glyph “error-->”. For example:

     $ echo hi on stdout
     -| hi on stdout
     $ echo hello on stderr 1>&2
     error--> hello on stderr

In the text, command names appear in this font, while code segments appear in the same font and quoted, ‘like this’. Some things are emphasized like this, and if a point needs to be made strongly, it is done like this. The first occurrence of a new term is usually its definition and appears in the same font as the previous occurrence of “definition” in this sentence. file names are indicated like this: /path/to/ourfile.

Characters that you type at the keyboard look like this. In particular, there are special characters called “control characters.” These are characters that you type by holding down both the CONTROL key and another key, at the same time. For example, a Ctrl-d is typed by first pressing and holding the CONTROL key, next pressing the d key and finally releasing both keys.

Dark Corners

Dark corners are basically fractal — no matter how much you illuminate, there's always a smaller but darker one.
Brian Kernighan

Until the POSIX standard (and The Gawk Manual), many features of awk were either poorly documented or not documented at all. Descriptions of such features (often called “dark corners”) are noted in this Web page with “(d.c.)”. They also appear in the index under the heading “dark corner.”

As noted by the opening quote, though, any coverage of dark corners is, by definition, something that is incomplete.

Next: , Previous: Conventions, Up: Preface

The GNU Project and This Book

The Free Software Foundation (FSF) is a nonprofit organization dedicated to the production and distribution of freely distributable software. It was founded by Richard M. Stallman, the author of the original Emacs editor. GNU Emacs is the most widely used version of Emacs today.

The GNU4 Project is an ongoing effort on the part of the Free Software Foundation to create a complete, freely distributable, POSIX-compliant computing environment. The FSF uses the “GNU General Public License” (GPL) to ensure that their software's source code is always available to the end user. A copy of the GPL is included in this Web page for your reference (see Copying). The GPL applies to the C language source code for gawk. To find out more about the FSF and the GNU Project online, see the GNU Project's home page. This Web page may also be read from their web site.

A shell, an editor (Emacs), highly portable optimizing C, C++, and Objective-C compilers, a symbolic debugger and dozens of large and small utilities (such as gawk), have all been completed and are freely available. The GNU operating system kernel (the HURD), has been released but is still in an early stage of development.

Until the GNU operating system is more fully developed, you should consider using GNU/Linux, a freely distributable, Unix-like operating system for Intel 80386, DEC Alpha, Sun SPARC, IBM S/390, and other systems.5 There are many books on GNU/Linux. One that is freely available is Linux Installation and Getting Started, by Matt Welsh. Many GNU/Linux distributions are often available in computer stores or bundled on CD-ROMs with books about Linux. (There are three other freely available, Unix-like operating systems for 80386 and other systems: NetBSD, FreeBSD, and OpenBSD. All are based on the 4.4-Lite Berkeley Software Distribution, and they use recent versions of gawk for their versions of awk.)

The Web page you are reading is actually free—at least, the information in it is free to anyone. The machine-readable source code for the Web page comes with gawk; anyone may take this Web page to a copying machine and make as many copies as they like. (Take a moment to check the Free Documentation License in GNU Free Documentation License.)

Although you could just print it out yourself, bound books are much easier to read and use. Furthermore, the proceeds from sales of this book go back to the FSF to help fund development of more free software.

The Web page itself has gone through a number of previous editions. Paul Rubin wrote the very first draft of The GAWK Manual; it was around 40 pages in size. Diane Close and Richard Stallman improved it, yielding a version that was around 90 pages long and barely described the original, “old” version of awk.

I started working with that version in the fall of 1988. As work on it progressed, the FSF published several preliminary versions (numbered 0.x). In 1996, Edition 1.0 was released with gawk 3.0.0. The FSF published the first two editions under the title The GNU Awk User's Guide.

This edition maintains the basic structure of Edition 1.0, but with significant additional material, reflecting the host of new features in gawk version 3.1. Of particular note is Array Sorting, as well as Bitwise Functions, Internationalization, and also Advanced Features, and Dynamic Extensions.

GAWK: Effective AWK Programming will undoubtedly continue to evolve. An electronic version comes with the gawk distribution from the FSF. If you find an error in this Web page, please report it! See Bugs, for information on submitting problem reports electronically, or write to me in care of the publisher.

Next: , Previous: Manual History, Up: Preface

How to Contribute

As the maintainer of GNU awk, I am starting a collection of publicly available awk programs. For more information, see If you have written an interesting awk program, or have written a gawk extension that you would like to share with the rest of the world, please contact me ( Making things available on the Internet helps keep the gawk distribution down to manageable size.

Previous: How To Contribute, Up: Preface


The initial draft of The GAWK Manual had the following acknowledgments:

Many people need to be thanked for their assistance in producing this manual. Jay Fenlason contributed many ideas and sample programs. Richard Mlynarik and Robert Chassell gave helpful comments on drafts of this manual. The paper A Supplemental Document for awk by John W. Pierce of the Chemistry Department at UC San Diego, pinpointed several issues relevant both to awk implementation and to this manual, that would otherwise have escaped us.

I would like to acknowledge Richard M. Stallman, for his vision of a better world and for his courage in founding the FSF and starting the GNU Project.

The following people (in alphabetical order) provided helpful comments on various versions of this book, up to and including this edition. Rick Adams, Nelson H.F. Beebe, Karl Berry, Dr. Michael Brennan, Rich Burridge, Claire Cloutier, Diane Close, Scott Deifik, Christopher (“Topher”) Eliot, Jeffrey Friedl, Dr. Darrel Hankerson, Michal Jaegermann, Dr. Richard J. LeBlanc, Michael Lijewski, Pat Rankin, Miriam Robbins, Mary Sheehan, and Chuck Toporek.

Robert J. Chassell provided much valuable advice on the use of Texinfo. He also deserves special thanks for convincing me not to title this Web page How To Gawk Politely. Karl Berry helped significantly with the TeX part of Texinfo.

I would like to thank Marshall and Elaine Hartholz of Seattle and Dr. Bert and Rita Schreiber of Detroit for large amounts of quiet vacation time in their homes, which allowed me to make significant progress on this Web page and on gawk itself.

Phil Hughes of SSC contributed in a very important way by loaning me his laptop GNU/Linux system, not once, but twice, which allowed me to do a lot of work while away from home.

David Trueman deserves special credit; he has done a yeoman job of evolving gawk so that it performs well and without bugs. Although he is no longer involved with gawk, working with him on this project was a significant pleasure.

The intrepid members of the GNITS mailing list, and most notably Ulrich Drepper, provided invaluable help and feedback for the design of the internationalization features.

Nelson Beebe, Martin Brown, Andreas Buening, Scott Deifik, Darrel Hankerson, Isamu Hasegawa, Michal Jaegermann, Jürgen Kahrs, Pat Rankin, Kai Uwe Rommel, and Eli Zaretskii (in alphabetical order) make up the gawk “crack portability team.” Without their hard work and help, gawk would not be nearly the fine program it is today. It has been and continues to be a pleasure working with this team of fine people.

David and I would like to thank Brian Kernighan of Bell Laboratories for invaluable assistance during the testing and debugging of gawk, and for help in clarifying numerous points about the language. We could not have done nearly as good a job on either gawk or its documentation without his help.

Chuck Toporek, Mary Sheehan, and Claire Coutier of O'Reilly & Associates contributed significant editorial help for this Web page for the 3.1 release of gawk.

I must thank my wonderful wife, Miriam, for her patience through the many versions of this project, for her proofreading, and for sharing me with the computer. I would like to thank my parents for their love, and for the grace with which they raised and educated me. Finally, I also must acknowledge my gratitude to G-d, for the many opportunities He has sent my way, as well as for the gifts He has given me with which to take advantage of those opportunities.

Arnold Robbins
Nof Ayalon
March, 2001

Next: , Previous: Preface, Up: Top

1 Getting Started with awk

The basic function of awk is to search files for lines (or other units of text) that contain certain patterns. When a line matches one of the patterns, awk performs specified actions on that line. awk keeps processing input lines in this way until it reaches the end of the input files.

Programs in awk are different from programs in most other languages, because awk programs are data-driven; that is, you describe the data you want to work with and then what to do when you find it. Most other languages are procedural; you have to describe, in great detail, every step the program is to take. When working with procedural languages, it is usually much harder to clearly describe the data your program will process. For this reason, awk programs are often refreshingly easy to read and write.

When you run awk, you specify an awk program that tells awk what to do. The program consists of a series of rules. (It may also contain function definitions, an advanced feature that we will ignore for now. See User-defined.) Each rule specifies one pattern to search for and one action to perform upon finding the pattern.

Syntactically, a rule consists of a pattern followed by an action. The action is enclosed in curly braces to separate it from the pattern. Newlines usually separate rules. Therefore, an awk program looks like this:

     pattern { action }
     pattern { action }

Next: , Up: Getting Started

1.1 How to Run awk Programs

There are several ways to run an awk program. If the program is short, it is easiest to include it in the command that runs awk, like this:

     awk 'program' input-file1 input-file2 ...

When the program is long, it is usually more convenient to put it in a file and run it with a command like this:

     awk -f program-file input-file1 input-file2 ...

This section discusses both mechanisms, along with several variations of each.

Next: , Up: Running gawk

1.1.1 One-Shot Throwaway awk Programs

Once you are familiar with awk, you will often type in simple programs the moment you want to use them. Then you can write the program as the first argument of the awk command, like this:

     awk 'program' input-file1 input-file2 ...

where program consists of a series of patterns and actions, as described earlier.

This command format instructs the shell, or command interpreter, to start awk and use the program to process records in the input file(s). There are single quotes around program so the shell won't interpret any awk characters as special shell characters. The quotes also cause the shell to treat all of program as a single argument for awk, and allow program to be more than one line long.

This format is also useful for running short or medium-sized awk programs from shell scripts, because it avoids the need for a separate file for the awk program. A self-contained shell script is more reliable because there are no other files to misplace.

Very Simple, later in this chapter, presents several short, self-contained programs.

Next: , Previous: One-shot, Up: Running gawk

1.1.2 Running awk Without Input Files

You can also run awk without any input files. If you type the following command line:

     awk 'program'

awk applies the program to the standard input, which usually means whatever you type on the terminal. This continues until you indicate end-of-file by typing Ctrl-d. (On other operating systems, the end-of-file character may be different. For example, on OS/2 and MS-DOS, it is Ctrl-z.)

As an example, the following program prints a friendly piece of advice (from Douglas Adams's The Hitchhiker's Guide to the Galaxy), to keep you from worrying about the complexities of computer programming (BEGIN is a feature we haven't discussed yet):

     $ awk "BEGIN { print \"Don't Panic!\" }"
     -| Don't Panic!

This program does not read any input. The ‘\’ before each of the inner double quotes is necessary because of the shell's quoting rules—in particular because it mixes both single quotes and double quotes.6

This next simple awk program emulates the cat utility; it copies whatever you type on the keyboard to its standard output (why this works is explained shortly).

     $ awk '{ print }'
     Now is the time for all good men
     -| Now is the time for all good men
     to come to the aid of their country.
     -| to come to the aid of their country.
     Four score and seven years ago, ...
     -| Four score and seven years ago, ...
     What, me worry?
     -| What, me worry?

Next: , Previous: Read Terminal, Up: Running gawk

1.1.3 Running Long Programs

Sometimes your awk programs can be very long. In this case, it is more convenient to put the program into a separate file. In order to tell awk to use that file for its program, you type:

     awk -f source-file input-file1 input-file2 ...

The -f instructs the awk utility to get the awk program from the file source-file. Any file name can be used for source-file. For example, you could put the program:

     BEGIN { print "Don't Panic!" }

into the file advice. Then this command:

     awk -f advice

does the same thing as this one:

     awk "BEGIN { print \"Don't Panic!\" }"

This was explained earlier (see Read Terminal). Note that you don't usually need single quotes around the file name that you specify with -f, because most file names don't contain any of the shell's special characters. Notice that in advice, the awk program did not have single quotes around it. The quotes are only needed for programs that are provided on the awk command line.

If you want to identify your awk program files clearly as such, you can add the extension .awk to the file name. This doesn't affect the execution of the awk program but it does make “housekeeping” easier.

Next: , Previous: Long, Up: Running gawk

1.1.4 Executable awk Programs

Once you have learned awk, you may want to write self-contained awk scripts, using the ‘#!’ script mechanism. You can do this on many Unix systems7 as well as on the GNU system. For example, you could update the file advice to look like this:

     #! /bin/awk -f
     BEGIN { print "Don't Panic!" }

After making this file executable (with the chmod utility), simply type ‘advice’ at the shell and the system arranges to run awk8 as if you had typed ‘awk -f advice’:

     $ chmod +x advice
     $ advice
     -| Don't Panic!

(We assume you have the current directory in your shell's search path variable (typically $PATH). If not, you may need to type ‘./advice’ at the shell.)

Self-contained awk scripts are useful when you want to write a program that users can invoke without their having to know that the program is written in awk.

Advanced Notes: Portability Issues with ‘#!

Some systems limit the length of the interpreter name to 32 characters. Often, this can be dealt with by using a symbolic link.

You should not put more than one argument on the ‘#!’ line after the path to awk. It does not work. The operating system treats the rest of the line as a single argument and passes it to awk. Doing this leads to confusing behavior—most likely a usage diagnostic of some sort from awk.

Finally, the value of ARGV[0] (see Built-in Variables) varies depending upon your operating system. Some systems put ‘awk’ there, some put the full pathname of awk (such as /bin/awk), and some put the name of your script (‘advice’). Don't rely on the value of ARGV[0] to provide your script name.

Next: , Previous: Executable Scripts, Up: Running gawk

1.1.5 Comments in awk Programs

A comment is some text that is included in a program for the sake of human readers; it is not really an executable part of the program. Comments can explain what the program does and how it works. Nearly all programming languages have provisions for comments, as programs are typically hard to understand without them.

In the awk language, a comment starts with the sharp sign character (‘#’) and continues to the end of the line. The ‘#’ does not have to be the first character on the line. The awk language ignores the rest of a line following a sharp sign. For example, we could have put the following into advice:

     # This program prints a nice friendly message.  It helps
     # keep novice users from being afraid of the computer.
     BEGIN    { print "Don't Panic!" }

You can put comment lines into keyboard-composed throwaway awk programs, but this usually isn't very useful; the purpose of a comment is to help you or another person understand the program when reading it at a later time.

Caution: As mentioned in One-shot, you can enclose small to medium programs in single quotes, in order to keep your shell scripts self-contained. When doing so, don't put an apostrophe (i.e., a single quote) into a comment (or anywhere else in your program). The shell interprets the quote as the closing quote for the entire program. As a result, usually the shell prints a message about mismatched quotes, and if awk actually runs, it will probably print strange messages about syntax errors. For example, look at the following:

     $ awk '{ print "hello" } # let's be cute'

The shell sees that the first two quotes match, and that a new quoted object begins at the end of the command line. It therefore prompts with the secondary prompt, waiting for more input. With Unix awk, closing the quoted string produces this result:

     $ awk '{ print "hello" } # let's be cute'
     > '
     error--> awk: can't open file be
     error-->  source line number 1

Putting a backslash before the single quote in ‘let's’ wouldn't help, since backslashes are not special inside single quotes. The next subsection describes the shell's quoting rules.

Previous: Comments, Up: Running gawk

1.1.6 Shell-Quoting Issues

For short to medium length awk programs, it is most convenient to enter the program on the awk command line. This is best done by enclosing the entire program in single quotes. This is true whether you are entering the program interactively at the shell prompt, or writing it as part of a larger shell script:

     awk 'program text' input-file1 input-file2 ...

Once you are working with the shell, it is helpful to have a basic knowledge of shell quoting rules. The following rules apply only to POSIX-compliant, Bourne-style shells (such as bash, the GNU Bourne-Again Shell). If you use csh, you're on your own.

Mixing single and double quotes is difficult. You have to resort to shell quoting tricks, like this:

     $ awk 'BEGIN { print "Here is a single quote <'"'"'>" }'
     -| Here is a single quote <'>

This program consists of three concatenated quoted strings. The first and the third are single-quoted, the second is double-quoted.

This can be “simplified” to:

     $ awk 'BEGIN { print "Here is a single quote <'\''>" }'
     -| Here is a single quote <'>

Judge for yourself which of these two is the more readable.

Another option is to use double quotes, escaping the embedded, awk-level double quotes:

     $ awk "BEGIN { print \"Here is a single quote <'>\" }"
     -| Here is a single quote <'>

This option is also painful, because double quotes, backslashes, and dollar signs are very common in awk programs.

A third option is to use the octal escape sequence equivalents for the single- and double-quote characters, like so:

     $ awk 'BEGIN { print "Here is a single quote <\47>" }'
     -| Here is a single quote <'>
     $ awk 'BEGIN { print "Here is a double quote <\42>" }'
     -| Here is a double quote <">

This works nicely, except that you should comment clearly what the escapes mean.

A fourth option is to use command-line variable assignment, like this:

     $ awk -v sq="'" 'BEGIN { print "Here is a single quote <" sq ">" }'
     -| Here is a single quote <'>

If you really need both single and double quotes in your awk program, it is probably best to move it into a separate file, where the shell won't be part of the picture, and you can say what you mean.

Next: , Previous: Running gawk, Up: Getting Started

1.2 Data Files for the Examples

Many of the examples in this Web page take their input from two sample data files. The first, BBS-list, represents a list of computer bulletin board systems together with information about those systems. The second data file, called inventory-shipped, contains information about monthly shipments. In both files, each line is considered to be one record.

In the data file BBS-list, each record contains the name of a computer bulletin board, its phone number, the board's baud rate(s), and a code for the number of hours it is operational. An ‘A’ in the last column means the board operates 24 hours a day. A ‘B’ in the last column means the board only operates on evening and weekend hours. A ‘C’ means the board operates only on weekends:

     aardvark     555-5553     1200/300          B
     alpo-net     555-3412     2400/1200/300     A
     barfly       555-7685     1200/300          A
     bites        555-1675     2400/1200/300     A
     camelot      555-0542     300               C
     core         555-2912     1200/300          C
     fooey        555-1234     2400/1200/300     B
     foot         555-6699     1200/300          B
     macfoo       555-6480     1200/300          A
     sdace        555-3430     2400/1200/300     A
     sabafoo      555-2127     1200/300          C

The data file inventory-shipped represents information about shipments during the year. Each record contains the month, the number of green crates shipped, the number of red boxes shipped, the number of orange bags shipped, and the number of blue packages shipped, respectively. There are 16 entries, covering the 12 months of last year and the first four months of the current year.

     Jan  13  25  15 115
     Feb  15  32  24 226
     Mar  15  24  34 228
     Apr  31  52  63 420
     May  16  34  29 208
     Jun  31  42  75 492
     Jul  24  34  67 436
     Aug  15  34  47 316
     Sep  13  55  37 277
     Oct  29  54  68 525
     Nov  20  87  82 577
     Dec  17  35  61 401
     Jan  21  36  64 620
     Feb  26  58  80 652
     Mar  24  75  70 495
     Apr  21  70  74 514

Next: , Previous: Sample Data Files, Up: Getting Started

1.3 Some Simple Examples

The following command runs a simple awk program that searches the input file BBS-list for the character string ‘foo’ (a grouping of characters is usually called a string; the term string is based on similar usage in English, such as “a string of pearls,” or “a string of cars in a train”):

     awk '/foo/ { print $0 }' BBS-list

When lines containing ‘foo’ are found, they are printed because ‘print $0 means print the current line. (Just ‘print’ by itself means the same thing, so we could have written that instead.)

You will notice that slashes (‘/’) surround the string ‘foo’ in the awk program. The slashes indicate that ‘foo’ is the pattern to search for. This type of pattern is called a regular expression, which is covered in more detail later (see Regexp). The pattern is allowed to match parts of words. There are single quotes around the awk program so that the shell won't interpret any of it as special shell characters.

Here is what this program prints:

     $ awk '/foo/ { print $0 }' BBS-list
     -| fooey        555-1234     2400/1200/300     B
     -| foot         555-6699     1200/300          B
     -| macfoo       555-6480     1200/300          A
     -| sabafoo      555-2127     1200/300          C

In an awk rule, either the pattern or the action can be omitted, but not both. If the pattern is omitted, then the action is performed for every input line. If the action is omitted, the default action is to print all lines that match the pattern.

Thus, we could leave out the action (the print statement and the curly braces) in the previous example and the result would be the same: all lines matching the pattern ‘foo’ are printed. By comparison, omitting the print statement but retaining the curly braces makes an empty action that does nothing (i.e., no lines are printed).

Many practical awk programs are just a line or two. Following is a collection of useful, short programs to get you started. Some of these programs contain constructs that haven't been covered yet. (The description of the program will give you a good idea of what is going on, but please read the rest of the Web page to become an awk expert!) Most of the examples use a data file named data. This is just a placeholder; if you use these programs yourself, substitute your own file names for data. For future reference, note that there is often more than one way to do things in awk. At some point, you may want to look back at these examples and see if you can come up with different ways to do the same things shown here:

Next: , Previous: Very Simple, Up: Getting Started

1.4 An Example with Two Rules

The awk utility reads the input files one line at a time. For each line, awk tries the patterns of each of the rules. If several patterns match, then several actions are run in the order in which they appear in the awk program. If no patterns match, then no actions are run.

After processing all the rules that match the line (and perhaps there are none), awk reads the next line. (However, see Next Statement, and also see Nextfile Statement). This continues until the program reaches the end of the file. For example, the following awk program contains two rules:

     /12/  { print $0 }
     /21/  { print $0 }

The first rule has the string ‘12’ as the pattern and ‘print $0’ as the action. The second rule has the string ‘21’ as the pattern and also has ‘print $0’ as the action. Each rule's action is enclosed in its own pair of braces.

This program prints every line that contains the string ‘12or the string ‘21’. If a line contains both strings, it is printed twice, once by each rule.

This is what happens if we run this program on our two sample data files, BBS-list and inventory-shipped:

     $ awk '/12/ { print $0 }
     >      /21/ { print $0 }' BBS-list inventory-shipped
     -| aardvark     555-5553     1200/300          B
     -| alpo-net     555-3412     2400/1200/300     A
     -| barfly       555-7685     1200/300          A
     -| bites        555-1675     2400/1200/300     A
     -| core         555-2912     1200/300          C
     -| fooey        555-1234     2400/1200/300     B
     -| foot         555-6699     1200/300          B
     -| macfoo       555-6480     1200/300          A
     -| sdace        555-3430     2400/1200/300     A
     -| sabafoo      555-2127     1200/300          C
     -| sabafoo      555-2127     1200/300          C
     -| Jan  21  36  64 620
     -| Apr  21  70  74 514

Note how the line beginning with ‘sabafoo’ in BBS-list was printed twice, once for each rule.

Next: , Previous: Two Rules, Up: Getting Started

1.5 A More Complex Example

Now that we've mastered some simple tasks, let's look at what typical awk programs do. This example shows how awk can be used to summarize, select, and rearrange the output of another utility. It uses features that haven't been covered yet, so don't worry if you don't understand all the details:

     ls -l | awk '$6 == "Nov" { sum += $5 }
                  END { print sum }'

This command prints the total number of bytes in all the files in the current directory that were last modified in November (of any year). 9 The ‘ls -l part of this example is a system command that gives you a listing of the files in a directory, including each file's size and the date the file was last modified. Its output looks like this:

     -rw-r--r--  1 arnold   user   1933 Nov  7 13:05 Makefile
     -rw-r--r--  1 arnold   user  10809 Nov  7 13:03 awk.h
     -rw-r--r--  1 arnold   user    983 Apr 13 12:14
     -rw-r--r--  1 arnold   user  31869 Jun 15 12:20 awkgram.y
     -rw-r--r--  1 arnold   user  22414 Nov  7 13:03 awk1.c
     -rw-r--r--  1 arnold   user  37455 Nov  7 13:03 awk2.c
     -rw-r--r--  1 arnold   user  27511 Dec  9 13:07 awk3.c
     -rw-r--r--  1 arnold   user   7989 Nov  7 13:03 awk4.c

The first field contains read-write permissions, the second field contains the number of links to the file, and the third field identifies the owner of the file. The fourth field identifies the group of the file. The fifth field contains the size of the file in bytes. The sixth, seventh, and eighth fields contain the month, day, and time, respectively, that the file was last modified. Finally, the ninth field contains the name of the file.10

The ‘$6 == "Nov"’ in our awk program is an expression that tests whether the sixth field of the output from ‘ls -l matches the string ‘Nov’. Each time a line has the string ‘Nov’ for its sixth field, the action ‘sum += $5’ is performed. This adds the fifth field (the file's size) to the variable sum. As a result, when awk has finished reading all the input lines, sum is the total of the sizes of the files whose lines matched the pattern. (This works because awk variables are automatically initialized to zero.)

After the last line of output from ls has been processed, the END rule executes and prints the value of sum. In this example, the value of sum is 80600.

These more advanced awk techniques are covered in later sections (see Action Overview). Before you can move on to more advanced awk programming, you have to know how awk interprets your input and displays your output. By manipulating fields and using print statements, you can produce some very useful and impressive-looking reports.

Next: , Previous: More Complex, Up: Getting Started

1.6 awk Statements Versus Lines

Most often, each line in an awk program is a separate statement or separate rule, like this:

     awk '/12/  { print $0 }
          /21/  { print $0 }' BBS-list inventory-shipped

However, gawk ignores newlines after any of the following symbols and keywords:

     ,    {    ?    :    ||    &&    do    else

A newline at any other point is considered the end of the statement.11

If you would like to split a single statement into two lines at a point where a newline would terminate it, you can continue it by ending the first line with a backslash character (‘\’). The backslash must be the final character on the line in order to be recognized as a continuation character. A backslash is allowed anywhere in the statement, even in the middle of a string or regular expression. For example:

     awk '/This regular expression is too long, so continue it\
      on the next line/ { print $1 }'

We have generally not used backslash continuation in the sample programs in this Web page. In gawk, there is no limit on the length of a line, so backslash continuation is never strictly necessary; it just makes programs more readable. For this same reason, as well as for clarity, we have kept most statements short in the sample programs presented throughout the Web page. Backslash continuation is most useful when your awk program is in a separate source file instead of entered from the command line. You should also note that many awk implementations are more particular about where you may use backslash continuation. For example, they may not allow you to split a string constant using backslash continuation. Thus, for maximum portability of your awk programs, it is best not to split your lines in the middle of a regular expression or a string.

Caution: Backslash continuation does not work as described with the C shell. It works for awk programs in files and for one-shot programs, provided you are using a POSIX-compliant shell, such as the Unix Bourne shell or bash. But the C shell behaves differently! There, you must use two backslashes in a row, followed by a newline. Note also that when using the C shell, every newline in your awk program must be escaped with a backslash. To illustrate:

     % awk 'BEGIN { \
     ?   print \\
     ?       "hello, world" \
     ? }'
     -| hello, world

Here, the ‘%’ and ‘?’ are the C shell's primary and secondary prompts, analogous to the standard shell's ‘$’ and ‘>’.

Compare the previous example to how it is done with a POSIX-compliant shell:

     $ awk 'BEGIN {
     >   print \
     >       "hello, world"
     > }'
     -| hello, world

awk is a line-oriented language. Each rule's action has to begin on the same line as the pattern. To have the pattern and action on separate lines, you must use backslash continuation; there is no other option.

Another thing to keep in mind is that backslash continuation and comments do not mix. As soon as awk sees the ‘#’ that starts a comment, it ignores everything on the rest of the line. For example:

     $ gawk 'BEGIN { print "dont panic" # a friendly \
     >                                    BEGIN rule
     > }'
     error--> gawk: cmd. line:2:                BEGIN rule
     error--> gawk: cmd. line:2:                ^ parse error

In this case, it looks like the backslash would continue the comment onto the next line. However, the backslash-newline combination is never even noticed because it is “hidden” inside the comment. Thus, the BEGIN is noted as a syntax error.

When awk statements within one rule are short, you might want to put more than one of them on a line. This is accomplished by separating the statements with a semicolon (‘;’). This also applies to the rules themselves. Thus, the program shown at the start of this section could also be written this way:

     /12/ { print $0 } ; /21/ { print $0 }
NOTE: The requirement that states that rules on the same line must be separated with a semicolon was not in the original awk language; it was added for consistency with the treatment of statements within an action.

Next: , Previous: Statements/Lines, Up: Getting Started

1.7 Other Features of awk

The awk language provides a number of predefined, or built-in, variables that your programs can use to get information from awk. There are other variables your program can set as well to control how awk processes your data.

In addition, awk provides a number of built-in functions for doing common computational and string-related operations. gawk provides built-in functions for working with timestamps, performing bit manipulation, and for runtime string translation.

As we develop our presentation of the awk language, we introduce most of the variables and many of the functions. They are defined systematically in Built-in Variables, and Built-in.

Previous: Other Features, Up: Getting Started

1.8 When to Use awk

Now that you've seen some of what awk can do, you might wonder how awk could be useful for you. By using utility programs, advanced patterns, field separators, arithmetic statements, and other selection criteria, you can produce much more complex output. The awk language is very useful for producing reports from large amounts of raw data, such as summarizing information from the output of other utility programs like ls. (See More Complex.)

Programs written with awk are usually much smaller than they would be in other languages. This makes awk programs easy to compose and use. Often, awk programs can be quickly composed at your terminal, used once, and thrown away. Because awk programs are interpreted, you can avoid the (usually lengthy) compilation part of the typical edit-compile-test-debug cycle of software development.

Complex programs have been written in awk, including a complete retargetable assembler for eight-bit microprocessors (see Glossary, for more information), and a microcode assembler for a special-purpose Prolog computer. More recently, gawk was used for writing a Wiki clone.12 While the original awk's capabilities were strained by tasks of such complexity, modern versions are more capable. Even the Bell Labs version of awk has fewer predefined limits, and those that it has are much larger than they used to be.

If you find yourself writing awk scripts of more than, say, a few hundred lines, you might consider using a different programming language. Emacs Lisp is a good choice if you need sophisticated string or pattern matching capabilities. The shell is also good at string and pattern matching; in addition, it allows powerful use of the system utilities. More conventional languages, such as C, C++, and Java, offer better facilities for system programming and for managing the complexity of large programs. Programs in these languages may require more lines of source code than the equivalent awk programs, but they are easier to maintain and usually run more efficiently.

Next: , Previous: Getting Started, Up: Top

2 Regular Expressions

A regular expression, or regexp, is a way of describing a set of strings. Because regular expressions are such a fundamental part of awk programming, their format and use deserve a separate chapter.

A regular expression enclosed in slashes (‘/’) is an awk pattern that matches every input record whose text belongs to that set. The simplest regular expression is a sequence of letters, numbers, or both. Such a regexp matches any string that contains that sequence. Thus, the regexp ‘foo’ matches any string containing ‘foo’. Therefore, the pattern /foo/ matches any input record containing the three characters ‘fooanywhere in the record. Other kinds of regexps let you specify more complicated classes of strings.

Initially, the examples in this chapter are simple. As we explain more about how regular expressions work, we will present more complicated instances.

Next: , Up: Regexp

2.1 How to Use Regular Expressions

A regular expression can be used as a pattern by enclosing it in slashes. Then the regular expression is tested against the entire text of each record. (Normally, it only needs to match some part of the text in order to succeed.) For example, the following prints the second field of each record that contains the string ‘foo’ anywhere in it:

     $ awk '/foo/ { print $2 }' BBS-list
     -| 555-1234
     -| 555-6699
     -| 555-6480
     -| 555-2127

~ (tilde), ~ operator Regular expressions can also be used in matching expressions. These expressions allow you to specify the string to match against; it need not be the entire current input record. The two operators ‘~’ and ‘!~’ perform regular expression comparisons. Expressions using these operators can be used as patterns, or in if, while, for, and do statements. (See Statements.) For example:

     exp ~ /regexp/

is true if the expression exp (taken as a string) matches regexp. The following example matches, or selects, all input records with the uppercase letter ‘J’ somewhere in the first field:

     $ awk '$1 ~ /J/' inventory-shipped
     -| Jan  13  25  15 115
     -| Jun  31  42  75 492
     -| Jul  24  34  67 436
     -| Jan  21  36  64 620

So does this:

     awk '{ if ($1 ~ /J/) print }' inventory-shipped

This next example is true if the expression exp (taken as a character string) does not match regexp:

     exp !~ /regexp/

The following example matches, or selects, all input records whose first field does not contain the uppercase letter ‘J’:

     $ awk '$1 !~ /J/' inventory-shipped
     -| Feb  15  32  24 226
     -| Mar  15  24  34 228
     -| Apr  31  52  63 420
     -| May  16  34  29 208

When a regexp is enclosed in slashes, such as /foo/, we call it a regexp constant, much like 5.27 is a numeric constant and "foo" is a string constant.

Next: , Previous: Regexp Usage, Up: Regexp

2.2 Escape Sequences

Some characters cannot be included literally in string constants ("foo") or regexp constants (/foo/). Instead, they should be represented with escape sequences, which are character sequences beginning with a backslash (‘\’). One use of an escape sequence is to include a double-quote character in a string constant. Because a plain double quote ends the string, you must use ‘\"’ to represent an actual double-quote character as a part of the string. For example:

     $ awk 'BEGIN { print "He said \"hi!\" to her." }'
     -| He said "hi!" to her.

The backslash character itself is another character that cannot be included normally; you must write ‘\\’ to put one backslash in the string or regexp. Thus, the string whose contents are the two characters ‘"’ and ‘\’ must be written "\"\\".

Backslash also represents unprintable characters such as TAB or newline. While there is nothing to stop you from entering most unprintable characters directly in a string constant or regexp constant, they may look ugly.

The following table lists all the escape sequences used in awk and what they represent. Unless noted otherwise, all these escape sequences apply to both string constants and regexp constants:

A literal backslash, ‘\’.

The “alert” character, Ctrl-g, ASCII code 7 (BEL). (This usually makes some sort of audible noise.)

Backspace, Ctrl-h, ASCII code 8 (BS).

Formfeed, Ctrl-l, ASCII code 12 (FF).

Newline, Ctrl-j, ASCII code 10 (LF).

Carriage return, Ctrl-m, ASCII code 13 (CR).

Horizontal TAB, Ctrl-i, ASCII code 9 (HT).

Vertical tab, Ctrl-k, ASCII code 11 (VT).

The octal value nnn, where nnn stands for 1 to 3 digits between ‘0’ and ‘7’. For example, the code for the ASCII ESC (escape) character is ‘\033’.

The hexadecimal value hh, where hh stands for a sequence of hexadecimal digits (‘0’–‘9’, and either ‘A’–‘F’ or ‘a’–‘f’). Like the same construct in ISO C, the escape sequence continues until the first nonhexadecimal digit is seen. However, using more than two hexadecimal digits produces undefined results. (The ‘\x’ escape sequence is not allowed in POSIX awk.)

A literal slash (necessary for regexp constants only). This expression is used when you want to write a regexp constant that contains a slash. Because the regexp is delimited by slashes, you need to escape the slash that is part of the pattern, in order to tell awk to keep processing the rest of the regexp.

A literal double quote (necessary for string constants only). This expression is used when you want to write a string constant that contains a double quote. Because the string is delimited by double quotes, you need to escape the quote that is part of the string, in order to tell awk to keep processing the rest of the string.

In gawk, a number of additional two-character sequences that begin with a backslash have special meaning in regexps. See GNU Regexp Operators.

In a regexp, a backslash before any character that is not in the previous list and not listed in GNU Regexp Operators, means that the next character should be taken literally, even if it would normally be a regexp operator. For example, /a\+b/ matches the three characters ‘a+b’.

For complete portability, do not use a backslash before any character not shown in the previous list.

To summarize:

Advanced Notes: Backslash Before Regular Characters

If you place a backslash in a string constant before something that is not one of the characters previously listed, POSIX awk purposely leaves what happens as undefined. There are two choices:

Strip the backslash out
This is what Unix awk and gawk both do. For example, "a\qc" is the same as "aqc". (Because this is such an easy bug both to introduce and to miss, gawk warns you about it.) Consider ‘FS = "[ \t]+\|[ \t]+"’ to use vertical bars surrounded by whitespace as the field separator. There should be two backslashes in the string ‘FS = "[ \t]+\\|[ \t]+"’.)

Leave the backslash alone
Some other awk implementations do this. In such implementations, typing "a\qc" is the same as typing "a\\qc".

Advanced Notes: Escape Sequences for Metacharacters

Suppose you use an octal or hexadecimal escape to represent a regexp metacharacter. (See Regexp Operators.) Does awk treat the character as a literal character or as a regexp operator?

Historically, such characters were taken literally. (d.c.) However, the POSIX standard indicates that they should be treated as real metacharacters, which is what gawk does. In compatibility mode (see Options), gawk treats the characters represented by octal and hexadecimal escape sequences literally when used in regexp constants. Thus, /a\52b/ is equivalent to /a\*b/.

Next: , Previous: Escape Sequences, Up: Regexp

2.3 Regular Expression Operators

You can combine regular expressions with special characters, called regular expression operators or metacharacters, to increase the power and versatility of regular expressions.

The escape sequences described earlier in Escape Sequences, are valid inside a regexp. They are introduced by a ‘\’ and are recognized and converted into corresponding real characters as the very first step in processing regexps.

Here is a list of metacharacters. All characters that are not escape sequences and that are not listed in the table stand for themselves:

This is used to suppress the special meaning of a character when matching. For example, ‘\$’ matches the character ‘$’.

This matches the beginning of a string. For example, ‘^@chapter’ matches ‘@chapter’ at the beginning of a string and can be used to identify chapter beginnings in Texinfo source files. The ‘^’ is known as an anchor, because it anchors the pattern to match only at the beginning of the string.

It is important to realize that ‘^’ does not match the beginning of a line embedded in a string. The condition is not true in the following example:

          if ("line1\nLINE 2" ~ /^L/) ...

This is similar to ‘^’, but it matches only at the end of a string. For example, ‘p$’ matches a record that ends with a ‘p’. The ‘$’ is an anchor and does not match the end of a line embedded in a string. The condition in the following example is not true:
          if ("line1\nLINE 2" ~ /1$/) ...

This matches any single character, including the newline character. For example, ‘.P’ matches any single character followed by a ‘P’ in a string. Using concatenation, we can make a regular expression such as ‘U.A’, which matches any three-character sequence that begins with ‘U’ and ends with ‘A’.

In strict POSIX mode (see Options), ‘.’ does not match the nul character, which is a character with all bits equal to zero. Otherwise, nul is just another character. Other versions of awk may not be able to match the nul character.

This is called a character list.13 It matches any one of the characters that are enclosed in the square brackets. For example, ‘[MVX]’ matches any one of the characters ‘M’, ‘V’, or ‘X’ in a string. A full discussion of what can be inside the square brackets of a character list is given in Character Lists.

[^ ...]
This is a complemented character list. The first character after the ‘[must be a ‘^’. It matches any characters except those in the square brackets. For example, ‘[^awk]’ matches any character that is not an ‘a’, ‘w’, or ‘k’.

This is the alternation operator and it is used to specify alternatives. The ‘|’ has the lowest precedence of all the regular expression operators. For example, ‘^P|[[:digit:]]’ matches any string that matches either ‘^P’ or ‘[[:digit:]]’. This means it matches any string that starts with ‘P’ or contains a digit.

The alternation applies to the largest possible regexps on either side.

Parentheses are used for grouping in regular expressions, as in arithmetic. They can be used to concatenate regular expressions containing the alternation operator, ‘|’. For example, ‘@(samp|code)\{[^}]+\}’ matches both ‘@code{foo}’ and ‘@samp{bar}’. (These are Texinfo formatting control sequences. The ‘+’ is explained further on in this list.)

This symbol means that the preceding regular expression should be repeated as many times as necessary to find a match. For example, ‘ph*’ applies the ‘*’ symbol to the preceding ‘h’ and looks for matches of one ‘p’ followed by any number of ‘h’s. This also matches just ‘p’ if no ‘h’s are present.

The ‘*’ repeats the smallest possible preceding expression. (Use parentheses if you want to repeat a larger expression.) It finds as many repetitions as possible. For example, ‘awk '/\(c[ad][ad]*r x\)/ { print }' sample’ prints every record in sample containing a string of the form ‘(car x)’, ‘(cdr x)’, ‘(cadr x)’, and so on. Notice the escaping of the parentheses by preceding them with backslashes.

This symbol is similar to ‘*’, except that the preceding expression must be matched at least once. This means that ‘wh+y’ would match ‘why’ and ‘whhy’, but not ‘wy’, whereas ‘wh*y’ would match all three of these strings. The following is a simpler way of writing the last ‘*’ example:
          awk '/\(c[ad]+r x\)/ { print }' sample

This symbol is similar to ‘*’, except that the preceding expression can be matched either once or not at all. For example, ‘fe?d’ matches ‘fed’ and ‘fd’, but nothing else.

One or two numbers inside braces denote an interval expression. If there is one number in the braces, the preceding regexp is repeated n times. If there are two numbers separated by a comma, the preceding regexp is repeated n to m times. If there is one number followed by a comma, then the preceding regexp is repeated at least n times:
Matches ‘whhhy’, but not ‘why’ or ‘whhhhy’.
Matches ‘whhhy’, ‘whhhhy’, or ‘whhhhhy’, only.
Matches ‘whhy’ or ‘whhhy’, and so on.

Interval expressions were not traditionally available in awk. They were added as part of the POSIX standard to make awk and egrep consistent with each other.

However, because old programs may use ‘{’ and ‘}’ in regexp constants, by default gawk does not match interval expressions in regexps. If either --posix or --re-interval are specified (see Options), then interval expressions are allowed in regexps.

For new programs that use ‘{’ and ‘}’ in regexp constants, it is good practice to always escape them with a backslash. Then the regexp constants are valid and work the way you want them to, using any version of awk.14

In regular expressions, the ‘*’, ‘+’, and ‘?’ operators, as well as the braces ‘{’ and ‘}’, have the highest precedence, followed by concatenation, and finally by ‘|’. As in arithmetic, parentheses can change how operators are grouped.

In POSIX awk and gawk, the ‘*’, ‘+’, and ‘?’ operators stand for themselves when there is nothing in the regexp that precedes them. For example, ‘/+/’ matches a literal plus sign. However, many other versions of awk treat such a usage as a syntax error.

If gawk is in compatibility mode (see Options), POSIX character classes and interval expressions are not available in regular expressions.

Next: , Previous: Regexp Operators, Up: Regexp

2.4 Using Character Lists

Within a character list, a range expression consists of two characters separated by a hyphen. It matches any single character that sorts between the two characters, using the locale's collating sequence and character set. For example, in the default C locale, ‘[a-dx-z]’ is equivalent to ‘[abcdxyz]’. Many locales sort characters in dictionary order, and in these locales, ‘[a-dx-z]’ is typically not equivalent to ‘[abcdxyz]’; instead it might be equivalent to ‘[aBbCcDdxXyYz]’, for example. To obtain the traditional interpretation of bracket expressions, you can use the C locale by setting the LC_ALL environment variable to the value ‘C’.

To include one of the characters ‘\’, ‘]’, ‘-’, or ‘^’ in a character list, put a ‘\’ in front of it. For example:


matches either ‘d’ or ‘]’.

This treatment of ‘\’ in character lists is compatible with other awk implementations and is also mandated by POSIX. The regular expressions in awk are a superset of the POSIX specification for Extended Regular Expressions (EREs). POSIX EREs are based on the regular expressions accepted by the traditional egrep utility.

Character classes are a new feature introduced in the POSIX standard. A character class is a special notation for describing lists of characters that have a specific attribute, but the actual characters can vary from country to country and/or from character set to character set. For example, the notion of what is an alphabetic character differs between the United States and France.

A character class is only valid in a regexp inside the brackets of a character list. Character classes consist of ‘[:’, a keyword denoting the class, and ‘:]’. table-char-classes lists the character classes defined by the POSIX standard.

Class Meaning
[:alnum:] Alphanumeric characters.
[:alpha:] Alphabetic characters.
[:blank:] Space and TAB characters.
[:cntrl:] Control characters.
[:digit:] Numeric characters.
[:graph:] Characters that are both printable and visible. (A space is printable but not visible, whereas an ‘a’ is both.)
[:lower:] Lowercase alphabetic characters.
[:print:] Printable characters (characters that are not control characters).
[:punct:] Punctuation characters (characters that are not letters, digits, control characters, or space characters).
[:space:] Space characters (such as space, TAB, and formfeed, to name a few).
[:upper:] Uppercase alphabetic characters.
[:xdigit:] Characters that are hexadecimal digits.

Table 2.1: POSIX Character Classes

For example, before the POSIX standard, you had to write /[A-Za-z0-9]/ to match alphanumeric characters. If your character set had other alphabetic characters in it, this would not match them, and if your character set collated differently from ASCII, this might not even match the ASCII alphanumeric characters. With the POSIX character classes, you can write /[[:alnum:]]/ to match the alphabetic and numeric characters in your character set.

Two additional special sequences can appear in character lists. These apply to non-ASCII character sets, which can have single symbols (called collating elements) that are represented with more than one character. They can also have several characters that are equivalent for collating, or sorting, purposes. (For example, in French, a plain “e” and a grave-accented “è” are equivalent.) These sequences are:

Collating symbols
Multicharacter collating elements enclosed between ‘[.’ and ‘.]’. For example, if ‘ch’ is a collating element, then [[.ch.]] is a regexp that matches this collating element, whereas [ch] is a regexp that matches either ‘c’ or ‘h’.

Equivalence classes
Locale-specific names for a list of characters that are equal. The name is enclosed between ‘[=’ and ‘=]’. For example, the name ‘e’ might be used to represent all of “e,” “è,” and “é.” In this case, [[=e=]] is a regexp that matches any of ‘e’, ‘é’, or ‘è’.

These features are very valuable in non-English-speaking locales.

Caution: The library functions that gawk uses for regular expression matching currently recognize only POSIX character classes; they do not recognize collating symbols or equivalence classes.

Next: , Previous: Character Lists, Up: Regexp

2.5 gawk-Specific Regexp Operators

GNU software that deals with regular expressions provides a number of additional regexp operators. These operators are described in this section and are specific to gawk; they are not available in other awk implementations. Most of the additional operators deal with word matching. For our purposes, a word is a sequence of one or more letters, digits, or underscores (‘_’):

Matches any word-constituent character—that is, it matches any letter, digit, or underscore. Think of it as shorthand for [[:alnum:]_].

Matches any character that is not word-constituent. Think of it as shorthand for [^[:alnum:]_].

Matches the empty string at the beginning of a word. For example, /\<away/ matches ‘away’ but not ‘stowaway’.

Matches the empty string at the end of a word. For example, /stow\>/ matches ‘stow’ but not ‘stowaway’.

Matches the empty string at either the beginning or the end of a word (i.e., the word boundary). For example, ‘\yballs?\y’ matches either ‘ball’ or ‘balls’, as a separate word.

Matches the empty string that occurs between two word-constituent characters. For example, /\Brat\B/ matches ‘crate’ but it does not match ‘dirty rat’. ‘\B’ is essentially the opposite of ‘\y’.

There are two other operators that work on buffers. In Emacs, a buffer is, naturally, an Emacs buffer. For other programs, gawk's regexp library routines consider the entire string to match as the buffer. The operators are:

Matches the empty string at the beginning of a buffer (string).

Matches the empty string at the end of a buffer (string).

Because ‘^’ and ‘$’ always work in terms of the beginning and end of strings, these operators don't add any new capabilities for awk. They are provided for compatibility with other GNU software.

In other GNU software, the word-boundary operator is ‘\b’. However, that conflicts with the awk language's definition of ‘\b’ as backspace, so gawk uses a different letter. An alternative method would have been to require two backslashes in the GNU operators, but this was deemed too confusing. The current method of using ‘\y’ for the GNU ‘\b’ appears to be the lesser of two evils.

The various command-line options (see Options) control how gawk interprets characters in regexps:

No options
In the default case, gawk provides all the facilities of POSIX regexps and the previously described GNU regexp operators. GNU regexp operators described in Regexp Operators. However, interval expressions are not supported.
Only POSIX regexps are supported; the GNU operators are not special (e.g., ‘\w’ matches a literal ‘w’). Interval expressions are allowed.
Traditional Unix awk regexps are matched. The GNU operators are not special, interval expressions are not available, nor are the POSIX character classes ([[:alnum:]], etc.). Characters described by octal and hexadecimal escape sequences are treated literally, even if they represent regexp metacharacters. Also, gawk silently skips directories named on the command line.
Allow interval expressions in regexps, even if --traditional has been provided. (--posix automatically enables interval expressions, so --re-interval is redundant when --posix is is used.)

Next: , Previous: GNU Regexp Operators, Up: Regexp

2.6 Case Sensitivity in Matching

Case is normally significant in regular expressions, both when matching ordinary characters (i.e., not metacharacters) and inside character sets. Thus, a ‘w’ in a regular expression matches only a lowercase ‘w’ and not an uppercase ‘W’.

The simplest way to do a case-independent match is to use a character list—for example, ‘[Ww]’. However, this can be cumbersome if you need to use it often, and it can make the regular expressions harder to read. There are two alternatives that you might prefer.

One way to perform a case-insensitive match at a particular point in the program is to convert the data to a single case, using the tolower or toupper built-in string functions (which we haven't discussed yet; see String Functions). For example:

     tolower($1) ~ /foo/  { ... }

converts the first field to lowercase before matching against it. This works in any POSIX-compliant awk.

Another method, specific to gawk, is to set the variable IGNORECASE to a nonzero value (see Built-in Variables). When IGNORECASE is not zero, all regexp and string operations ignore case. Changing the value of IGNORECASE dynamically controls the case-sensitivity of the program as it runs. Case is significant by default because IGNORECASE (like most variables) is initialized to zero:

     x = "aB"
     if (x ~ /ab/) ...   # this test will fail
     if (x ~ /ab/) ...   # now it will succeed

In general, you cannot use IGNORECASE to make certain rules case-insensitive and other rules case-sensitive, because there is no straightforward way to set IGNORECASE just for the pattern of a particular rule.15 To do this, use either character lists or tolower. However, one thing you can do with IGNORECASE only is dynamically turn case-sensitivity on or off for all the rules at once.

IGNORECASE can be set on the command line or in a BEGIN rule (see Other Arguments; also see Using BEGIN/END). Setting IGNORECASE from the command line is a way to make a program case-insensitive without having to edit it.

Prior to gawk 3.0, the value of IGNORECASE affected regexp operations only. It did not affect string comparison with ‘==’, ‘!=’, and so on. Beginning with version 3.0, both regexp and string comparison operations are also affected by IGNORECASE.

Beginning with gawk 3.0, the equivalences between upper- and lowercase characters are based on the ISO-8859-1 (ISO Latin-1) character set. This character set is a superset of the traditional 128 ASCII characters, which also provides a number of characters suitable for use with European languages.

As of gawk 3.1.4, the case equivalences are fully locale-aware. They are based on the C <ctype.h> facilities, such as isalpha() and toupper().

The value of IGNORECASE has no effect if gawk is in compatibility mode (see Options). Case is always significant in compatibility mode.

Next: , Previous: Case-sensitivity, Up: Regexp

2.7 How Much Text Matches?

Consider the following:

     echo aaaabcd | awk '{ sub(/a+/, "<A>"); print }'

This example uses the sub function (which we haven't discussed yet; see String Functions) to make a change to the input record. Here, the regexp /a+/ indicates “one or more ‘a’ characters,” and the replacement text is ‘<A>’.

The input contains four ‘a’ characters. awk (and POSIX) regular expressions always match the leftmost, longest sequence of input characters that can match. Thus, all four ‘a’ characters are replaced with ‘<A>’ in this example:

     $ echo aaaabcd | awk '{ sub(/a+/, "<A>"); print }'
     -| <A>bcd

For simple match/no-match tests, this is not so important. But when doing text matching and substitutions with the match, sub, gsub, and gensub functions, it is very important. Understanding this principle is also important for regexp-based record and field splitting (see Records, and also see Field Separators).

Next: , Previous: Leftmost Longest, Up: Regexp

2.8 Using Dynamic Regexps

The righthand side of a ‘~’ or ‘!~’ operator need not be a regexp constant (i.e., a string of characters between slashes). It may be any expression. The expression is evaluated and converted to a string if necessary; the contents of the string are used as the regexp. A regexp that is computed in this way is called a dynamic regexp:

     BEGIN { digits_regexp = "[[:digit:]]+" }
     $0 ~ digits_regexp    { print }

This sets digits_regexp to a regexp that describes one or more digits, and tests whether the input record matches this regexp.

Caution: When using the ‘~’ and ‘!~’ operators, there is a difference between a regexp constant enclosed in slashes and a string constant enclosed in double quotes. If you are going to use a string constant, you have to understand that the string is, in essence, scanned twice: the first time when awk reads your program, and the second time when it goes to match the string on the lefthand side of the operator with the pattern on the right. This is true of any string-valued expression (such as digits_regexp, shown previously), not just string constants.

What difference does it make if the string is scanned twice? The answer has to do with escape sequences, and particularly with backslashes. To get a backslash into a regular expression inside a string, you have to type two backslashes.

For example, /\*/ is a regexp constant for a literal ‘*’. Only one backslash is needed. To do the same thing with a string, you have to type "\\*". The first backslash escapes the second one so that the string actually contains the two characters ‘\’ and ‘*’.

Given that you can use both regexp and string constants to describe regular expressions, which should you use? The answer is “regexp constants,” for several reasons:

Advanced Notes: Using \n in Character Lists of Dynamic Regexps

Some commercial versions of awk do not allow the newline character to be used inside a character list for a dynamic regexp:

     $ awk '$0 ~ "[ \t\n]"'
     error--> awk: newline in character class [
     error--> ]...
     error-->  source line number 1
     error-->  context is
     error-->          >>>  <<<

But a newline in a regexp constant works with no problem:

     $ awk '$0 ~ /[ \t\n]/'
     here is a sample line
     -| here is a sample line

gawk does not have this problem, and it isn't likely to occur often in practice, but it's worth noting for future reference.

Previous: Computed Regexps, Up: Regexp

2.9 Where You Are Makes A Difference

Modern systems support the notion of locales: a way to tell the system about the local character set and language. The current locale setting can affect the way regexp matching works, often in surprising ways. In particular, many locales do case-insensitive matching, even when you may have specified characters of only one particular case.

The following example uses the sub function, which does text replacement (see String Functions). Here, the intent is to remove trailing uppercase characters:

     $ echo something1234abc | gawk '{ sub("[A-Z]*$", ""); print }'
     -| something1234

This output is unexpected, since the ‘abc’ at the end of ‘something1234abc’ should not normally match ‘[A-Z]*’. This result is due to the locale setting (and thus you may not see it on your system). There are two fixes. The first is to use the POSIX character class ‘[[:upper:]]’, instead of ‘[A-Z]’. (This is preferred, since then your program will work everywhere.) The second is to change the locale setting in the environment, before running gawk, by using the shell statements:

     export LANG LC_ALL

The setting ‘C’ forces gawk to behave in the traditional Unix manner, where case distinctions do matter. You may wish to put these statements into your shell startup file, e.g., $HOME/.profile.

Similar considerations apply to other ranges. For example, ‘["-/]’ is perfectly valid in ASCII, but is not valid in many Unicode locales, such as ‘en_US.UTF-8’. (In general, such ranges should be avoided; either list the characters individually, or use a POSIX character class such as ‘[[:punct:]]’.)

For the normal case of ‘RS = "\n"’, the locale is largely irrelevant. For other single-character record separators, using ‘LC_ALL=C’ will give you much better performance when reading records. Otherwise, gawk has to make several function calls, per input character to find the record terminator.

Finally, the locale affects the value of the decimal point character used when gawk parses input data. This is discussed in detail in Conversion.

Next: , Previous: Regexp, Up: Top

3 Reading Input Files

In the typical awk program, all input is read either from the standard input (by default, this is the keyboard, but often it is a pipe from another command) or from files whose names you specify on the awk command line. If you specify input files, awk reads them in order, processing all the data from one before going on to the next. The name of the current input file can be found in the built-in variable FILENAME (see Built-in Variables).

The input is read in units called records, and is processed by the rules of your program one record at a time. By default, each record is one line. Each record is automatically split into chunks called fields. This makes it more convenient for programs to work on the parts of a record.

On rare occasions, you may need to use the getline command. The getline command is valuable, both because it can do explicit input from any number of files, and because the files used with it do not have to be named on the awk command line (see Getline).

Next: , Up: Reading Files

3.1 How Input Is Split into Records

The awk utility divides the input for your awk program into records and fields. awk keeps track of the number of records that have been read so far from the current input file. This value is stored in a built-in variable called FNR. It is reset to zero when a new file is started. Another built-in variable, NR, is the total number of input records read so far from all data files. It starts at zero, but is never automatically reset to zero.

Records are separated by a character called the record separator. By default, the record separator is the newline character. This is why records are, by default, single lines. A different character can be used for the record separator by assigning the character to the built-in variable RS.

Like any other variable, the value of RS can be changed in the awk program with the assignment operator, ‘=’ (see Assignment Ops). The new record-separator character should be enclosed in quotation marks, which indicate a string constant. Often the right time to do this is at the beginning of execution, before any input is processed, so that the very first record is read with the proper separator. To do this, use the special BEGIN pattern (see BEGIN/END). For example:

     awk 'BEGIN { RS = "/" }
          { print $0 }' BBS-list

changes the value of RS to "/", before reading any input. This is a string whose first character is a slash; as a result, records are separated by slashes. Then the input file is read, and the second rule in the awk program (the action with no pattern) prints each record. Because each print statement adds a newline at the end of its output, this awk program copies the input with each slash changed to a newline. Here are the results of running the program on BBS-list:

     $ awk 'BEGIN { RS = "/" }
     >      { print $0 }' BBS-list
     -| aardvark     555-5553     1200
     -| 300          B
     -| alpo-net     555-3412     2400
     -| 1200
     -| 300     A
     -| barfly       555-7685     1200
     -| 300          A
     -| bites        555-1675     2400
     -| 1200
     -| 300     A
     -| camelot      555-0542     300               C
     -| core         555-2912     1200
     -| 300          C
     -| fooey        555-1234     2400
     -| 1200
     -| 300     B
     -| foot         555-6699     1200
     -| 300          B
     -| macfoo       555-6480     1200
     -| 300          A
     -| sdace        555-3430     2400
     -| 1200
     -| 300     A
     -| sabafoo      555-2127     1200
     -| 300          C

Note that the entry for the ‘camelot’ BBS is not split. In the original data file (see Sample Data Files), the line looks like this:

     camelot      555-0542     300               C

It has one baud rate only, so there are no slashes in the record, unlike the others which have two or more baud rates. In fact, this record is treated as part of the record for the ‘core’ BBS; the newline separating them in the output is the original newline in the data file, not the one added by awk when it printed the record!

Another way to change the record separator is on the command line, using the variable-assignment feature (see Other Arguments):

     awk '{ print $0 }' RS="/" BBS-list

This sets RS to ‘/’ before processing BBS-list.

Using an unusual character such as ‘/’ for the record separator produces correct behavior in the vast majority of cases. However, the following (extreme) pipeline prints a surprising ‘1’:

     $ echo | awk 'BEGIN { RS = "a" } ; { print NF }'
     -| 1

There is one field, consisting of a newline. The value of the built-in variable NF is the number of fields in the current record.

Reaching the end of an input file terminates the current input record, even if the last character in the file is not the character in RS. (d.c.)

The empty string "" (a string without any characters) has a special meaning as the value of RS. It means that records are separated by one or more blank lines and nothing else. See Multiple Line, for more details.

If you change the value of RS in the middle of an awk run, the new value is used to delimit subsequent records, but the record currently being processed, as well as records already processed, are not affected.

After the end of the record has been determined, gawk sets the variable RT to the text in the input that matched RS. When using gawk, the value of RS is not limited to a one-character string. It can be any regular expression (see Regexp). In general, each record ends at the next string that matches the regular expression; the next record starts at the end of the matching string. This general rule is actually at work in the usual case, where RS contains just a newline: a record ends at the beginning of the next matching string (the next newline in the input), and the following record starts just after the end of this string (at the first character of the following line). The newline, because it matches RS, is not part of either record.

When RS is a single character, RT contains the same single character. However, when RS is a regular expression, RT contains the actual input text that matched the regular expression.

The following example illustrates both of these features. It sets RS equal to a regular expression that matches either a newline or a series of one or more uppercase letters with optional leading and/or trailing whitespace:

     $ echo record 1 AAAA record 2 BBBB record 3 |
     > gawk 'BEGIN { RS = "\n|( *[[:upper:]]+ *)" }
     >             { print "Record =", $0, "and RT =", RT }'
     -| Record = record 1 and RT =  AAAA
     -| Record = record 2 and RT =  BBBB
     -| Record = record 3 and RT =

The final line of output has an extra blank line. This is because the value of RT is a newline, and the print statement supplies its own terminating newline. See Simple Sed, for a more useful example of RS as a regexp and RT.

If you set RS to a regular expression that allows optional trailing text, such as ‘RS = "abc(XYZ)?"’ it is possible, due to implementation constraints, that gawk may match the leading part of the regular expression, but not the trailing part, particularly if the input text that could match the trailing part is fairly long. gawk attempts to avoid this problem, but currently, there's no guarantee that this will never happen.

NOTE: Remember that in awk, the ‘^’ and ‘$’ anchor metacharacters match the beginning and end of a string, and not the beginning and end of a line. As a result, something like ‘RS = "^[[:upper:]]"’ can only match at the beginning of a file. This is because gawk views the input file as one long string that happens to contain newline characters in it. It is thus best to avoid anchor characters in the value of RS.

The use of RS as a regular expression and the RT variable are gawk extensions; they are not available in compatibility mode (see Options). In compatibility mode, only the first character of the value of RS is used to determine the end of the record.

Advanced Notes: RS = "\0" Is Not Portable

There are times when you might want to treat an entire data file as a single record. The only way to make this happen is to give RS a value that you know doesn't occur in the input file. This is hard to do in a general way, such that a program always works for arbitrary input files.

You might think that for text files, the nul character, which consists of a character with all bits equal to zero, is a good value to use for RS in this case:

     BEGIN { RS = "\0" }  # whole file becomes one record?

gawk in fact accepts this, and uses the nul character for the record separator. However, this usage is not portable to other awk implementations.

All other awk implementations16 store strings internally as C-style strings. C strings use the nul character as the string terminator. In effect, this means that ‘RS = "\0"’ is the same as ‘RS = ""’. (d.c.)

The best way to treat a whole file as a single record is to simply read the file in, one record at a time, concatenating each record onto the end of the previous ones.

Next: , Previous: Records, Up: Reading Files

3.2 Examining Fields

When awk reads an input record, the record is automatically parsed or separated by the interpreter into chunks called fields. By default, fields are separated by whitespace, like words in a line. Whitespace in awk means any string of one or more spaces, tabs, or newlines;17 other characters, such as formfeed, vertical tab, etc. that are considered whitespace by other languages, are not considered whitespace by awk.

The purpose of fields is to make it more convenient for you to refer to these pieces of the record. You don't have to use them—you can operate on the whole record if you want—but fields are what make simple awk programs so powerful.

A dollar-sign (‘$’) is used to refer to a field in an awk program, followed by the number of the field you want. Thus, $1 refers to the first field, $2 to the second, and so on. (Unlike the Unix shells, the field numbers are not limited to single digits. $127 is the one hundred twenty-seventh field in the record.) For example, suppose the following is a line of input:

     This seems like a pretty nice example.

Here the first field, or $1, is ‘This’, the second field, or $2, is ‘seems’, and so on. Note that the last field, $7, is ‘example.’. Because there is no space between the ‘e’ and the ‘.’, the period is considered part of the seventh field.

NF is a built-in variable whose value is the number of fields in the current record. awk automatically updates the value of NF each time it reads a record. No matter how many fields there are, the last field in a record can be represented by $NF. So, $NF is the same as $7, which is ‘example.’. If you try to reference a field beyond the last one (such as $8 when the record has only seven fields), you get the empty string. (If used in a numeric operation, you get zero.)

The use of $0, which looks like a reference to the “zero-th” field, is a special case: it represents the whole input record when you are not interested in specific fields. Here are some more examples:

     $ awk '$1 ~ /foo/ { print $0 }' BBS-list
     -| fooey        555-1234     2400/1200/300     B
     -| foot         555-6699     1200/300          B
     -| macfoo       555-6480     1200/300          A
     -| sabafoo      555-2127     1200/300          C

This example prints each record in the file BBS-list whose first field contains the string ‘foo’. The operator ‘~’ is called a matching operator (see Regexp Usage); it tests whether a string (here, the field $1) matches a given regular expression.

By contrast, the following example looks for ‘foo’ in the entire record and prints the first field and the last field for each matching input record:

     $ awk '/foo/ { print $1, $NF }' BBS-list
     -| fooey B
     -| foot B
     -| macfoo A
     -| sabafoo C

Next: , Previous: Fields, Up: Reading Files

3.3 Nonconstant Field Numbers

The number of a field does not need to be a constant. Any expression in the awk language can be used after a ‘$’ to refer to a field. The value of the expression specifies the field number. If the value is a string, rather than a number, it is converted to a number. Consider this example:

     awk '{ print $NR }'

Recall that NR is the number of records read so far: one in the first record, two in the second, etc. So this example prints the first field of the first record, the second field of the second record, and so on. For the twentieth record, field number 20 is printed; most likely, the record has fewer than 20 fields, so this prints a blank line. Here is another example of using expressions as field numbers:

     awk '{ print $(2*2) }' BBS-list

awk evaluates the expression ‘(2*2)’ and uses its value as the number of the field to print. The ‘*’ sign represents multiplication, so the expression ‘2*2’ evaluates to four. The parentheses are used so that the multiplication is done before the ‘$’ operation; they are necessary whenever there is a binary operator in the field-number expression. This example, then, prints the hours of operation (the fourth field) for every line of the file BBS-list. (All of the awk operators are listed, in order of decreasing precedence, in Precedence.)

If the field number you compute is zero, you get the entire record. Thus, ‘$(2-2)’ has the same value as $0. Negative field numbers are not allowed; trying to reference one usually terminates the program. (The POSIX standard does not define what happens when you reference a negative field number. gawk notices this and terminates your program. Other awk implementations may behave differently.)

As mentioned in Fields, awk stores the current record's number of fields in the built-in variable NF (also see Built-in Variables). The expression $NF is not a special feature—it is the direct consequence of evaluating NF and using its value as a field number.

Next: , Previous: Nonconstant Fields, Up: Reading Files

3.4 Changing the Contents of a Field

The contents of a field, as seen by awk, can be changed within an awk program; this changes what awk perceives as the current input record. (The actual input is untouched; awk never modifies the input file.) Consider the following example and its output:

     $ awk '{ nboxes = $3 ; $3 = $3 - 10
     >        print nboxes, $3 }' inventory-shipped
     -| 25 15
     -| 32 22
     -| 24 14

The program first saves the original value of field three in the variable nboxes. The ‘-’ sign represents subtraction, so this program reassigns field three, $3, as the original value of field three minus ten: ‘$3 - 10’. (See Arithmetic Ops.) Then it prints the original and new values for field three. (Someone in the warehouse made a consistent mistake while inventorying the red boxes.)

For this to work, the text in field $3 must make sense as a number; the string of characters must be converted to a number for the computer to do arithmetic on it. The number resulting from the subtraction is converted back to a string of characters that then becomes field three. See Conversion.

When the value of a field is changed (as perceived by awk), the text of the input record is recalculated to contain the new field where the old one was. In other words, $0 changes to reflect the altered field. Thus, this program prints a copy of the input file, with 10 subtracted from the second field of each line:

     $ awk '{ $2 = $2 - 10; print $0 }' inventory-shipped
     -| Jan 3 25 15 115
     -| Feb 5 32 24 226
     -| Mar 5 24 34 228

It is also possible to also assign contents to fields that are out of range. For example:

     $ awk '{ $6 = ($5 + $4 + $3 + $2)
     >        print $6 }' inventory-shipped
     -| 168
     -| 297
     -| 301

We've just created $6, whose value is the sum of fields $2, $3, $4, and $5. The ‘+’ sign represents addition. For the file inventory-shipped, $6 represents the total number of parcels shipped for a particular month.

Creating a new field changes awk's internal copy of the current input record, which is the value of $0. Thus, if you do ‘print $0’ after adding a field, the record printed includes the new field, with the appropriate number of field separators between it and the previously existing fields.

This recomputation affects and is affected by NF (the number of fields; see Fields). For example, the value of NF is set to the number of the highest field you create. The exact format of $0 is also affected by a feature that has not been discussed yet: the output field separator, OFS, used to separate the fields (see Output Separators).

Note, however, that merely referencing an out-of-range field does not change the value of either $0 or NF. Referencing an out-of-range field only produces an empty string. For example:

     if ($(NF+1) != "")
         print "can't happen"
         print "everything is normal"

should print ‘everything is normal’, because NF+1 is certain to be out of range. (See If Statement, for more information about awk's if-else statements. See Typing and Comparison, for more information about the ‘!=’ operator.)

It is important to note that making an assignment to an existing field changes the value of $0 but does not change the value of NF, even when you assign the empty string to a field. For example:

     $ echo a b c d | awk '{ OFS = ":"; $2 = ""
     >                       print $0; print NF }'
     -| a::c:d
     -| 4

The field is still there; it just has an empty value, denoted by the two colons between ‘a’ and ‘c’. This example shows what happens if you create a new field:

     $ echo a b c d | awk '{ OFS = ":"; $2 = ""; $6 = "new"
     >                       print $0; print NF }'
     -| a::c:d::new
     -| 6

The intervening field, $5, is created with an empty value (indicated by the second pair of adjacent colons), and NF is updated with the value six.

Decrementing NF throws away the values of the fields after the new value of NF and recomputes $0. (d.c.) Here is an example:

     $ echo a b c d e f | awk '{ print "NF =", NF;
     >                            NF = 3; print $0 }'
     -| NF = 6
     -| a b c

Caution: Some versions of awk don't rebuild $0 when NF is decremented. Caveat emptor.

Finally, there are times when it is convenient to force awk to rebuild the entire record, using the current value of the fields and OFS. To do this, use the seemingly innocuous assignment:

     $1 = $1   # force record to be reconstituted
     print $0  # or whatever else with $0

This forces awk rebuild the record. It does help to add a comment, as we've shown here.

There is a flip side to the relationship between $0 and the fields. Any assignment to $0 causes the record to be reparsed into fields using the current value of FS. This also applies to any built-in function that updates $0, such as sub and gsub (see String Functions).

Next: , Previous: Changing Fields, Up: Reading Files

3.5 Specifying How Fields Are Separated

The field separator, which is either a single character or a regular expression, controls the way awk splits an input record into fields. awk scans the input record for character sequences that match the separator; the fields themselves are the text between the matches.

In the examples that follow, we use the bullet symbol (•) to represent spaces in the output. If the field separator is ‘oo’, then the following line:

     moo goo gai pan

is split into three fields: ‘m’, ‘•g’, and ‘•gai•pan’. Note the leading spaces in the values of the second and third fields.

The field separator is represented by the built-in variable FS. Shell programmers take note: awk does not use the name IFS that is used by the POSIX-compliant shells (such as the Unix Bourne shell, sh, or bash).

The value of FS can be changed in the awk program with the assignment operator, ‘=’ (see Assignment Ops). Often the right time to do this is at the beginning of execution before any input has been processed, so that the very first record is read with the proper separator. To do this, use the special BEGIN pattern (see BEGIN/END). For example, here we set the value of FS to the string ",":

     awk 'BEGIN { FS = "," } ; { print $2 }'

Given the input line:

     John Q. Smith, 29 Oak St., Walamazoo, MI 42139

this awk program extracts and prints the string ‘•29•Oak•St.’.

Sometimes the input data contains separator characters that don't separate fields the way you thought they would. For instance, the person's name in the example we just used might have a title or suffix attached, such as:

     John Q. Smith, LXIX, 29 Oak St., Walamazoo, MI 42139

The same program would extract ‘•LXIX’, instead of ‘•29•Oak•St.’. If you were expecting the program to print the address, you would be surprised. The moral is to choose your data layout and separator characters carefully to prevent such problems. (If the data is not in a form that is easy to process, perhaps you can massage it first with a separate awk program.)

Fields are normally separated by whitespace sequences (spaces, tabs, and newlines), not by single spaces. Two spaces in a row do not delimit an empty field. The default value of the field separator FS is a string containing a single space, " ". If awk interpreted this value in the usual way, each space character would separate fields, so two spaces in a row would make an empty field between them. The reason this does not happen is that a single space as the value of FS is a special case—it is taken to specify the default manner of delimiting fields.

If FS is any other single character, such as ",", then each occurrence of that character separates two fields. Two consecutive occurrences delimit an empty field. If the character occurs at the beginning or the end of the line, that too delimits an empty field. The space character is the only single character that does not follow these rules.

Next: , Up: Field Separators

3.5.1 Using Regular Expressions to Separate Fields

The previous subsection discussed the use of single characters or simple strings as the value of FS. More generally, the value of FS may be a string containing any regular expression. In this case, each match in the record for the regular expression separates fields. For example, the assignment:

     FS = ", \t"

makes every area of an input line that consists of a comma followed by a space and a TAB into a field separator.

For a less trivial example of a regular expression, try using single spaces to separate fields the way single commas are used. FS can be set to "[ ]" (left bracket, space, right bracket). This regular expression matches a single space and nothing else (see Regexp).

There is an important difference between the two cases of ‘FS = " "’ (a single space) and ‘FS = "[ \t\n]+"’ (a regular expression matching one or more spaces, tabs, or newlines). For both values of FS, fields are separated by runs (multiple adjacent occurrences) of spaces, tabs, and/or newlines. However, when the value of FS is " ", awk first strips leading and trailing whitespace from the record and then decides where the fields are. For example, the following pipeline prints ‘b’:

     $ echo ' a b c d ' | awk '{ print $2 }'
     -| b

However, this pipeline prints ‘a’ (note the extra spaces around each letter):

     $ echo ' a  b  c  d ' | awk 'BEGIN { FS = "[ \t\n]+" }
     >                                  { print $2 }'
     -| a

In this case, the first field is null or empty.

The stripping of leading and trailing whitespace also comes into play whenever $0 is recomputed. For instance, study this pipeline:

     $ echo '   a b c d' | awk '{ print; $2 = $2; print }'
     -|    a b c d
     -| a b c d

The first print statement prints the record as it was read, with leading whitespace intact. The assignment to $2 rebuilds $0 by concatenating $1 through $NF together, separated by the value of OFS. Because the leading whitespace was ignored when finding $1, it is not part of the new $0. Finally, the last print statement prints the new $0.

Next: , Previous: Regexp Field Splitting, Up: Field Separators

3.5.2 Making Each Character a Separate Field

There are times when you may want to examine each character of a record separately. This can be done in gawk by simply assigning the null string ("") to FS. In this case, each individual character in the record becomes a separate field. For example:

     $ echo a b | gawk 'BEGIN { FS = "" }
     >                  {
     >                      for (i = 1; i <= NF; i = i + 1)
     >                          print "Field", i, "is", $i
     >                  }'
     -| Field 1 is a
     -| Field 2 is
     -| Field 3 is b

Traditionally, the behavior of FS equal to "" was not defined. In this case, most versions of Unix awk simply treat the entire record as only having one field. (d.c.) In compatibility mode (see Options), if FS is the null string, then gawk also behaves this way.

Next: , Previous: Single Character Fields, Up: Field Separators

3.5.3 Setting FS from the Command Line

FS can be set on the command line. Use the -F option to do so. For example:

     awk -F, 'program' input-files

sets FS to the ‘,’ character. Notice that the option uses an uppercase ‘F’ instead of a lowercase ‘f’. The latter option (-f) specifies a file containing an awk program. Case is significant in command-line options: the -F and -f options have nothing to do with each other. You can use both options at the same time to set the FS variable and get an awk program from a file.

The value used for the argument to -F is processed in exactly the same way as assignments to the built-in variable FS. Any special characters in the field separator must be escaped appropriately. For example, to use a ‘\’ as the field separator on the command line, you would have to type:

     # same as FS = "\\"
     awk -F\\\\ '...' files ...

Because ‘\’ is used for quoting in the shell, awk sees ‘-F\\’. Then awk processes the ‘\\’ for escape characters (see Escape Sequences), finally yielding a single ‘\’ to use for the field separator.

As a special case, in compatibility mode (see Options), if the argument to -F is ‘t’, then FS is set to the TAB character. If you type ‘-F\t’ at the shell, without any quotes, the ‘\’ gets deleted, so awk figures that you really want your fields to be separated with tabs and not ‘t’s. Use ‘-v FS="t"’ or ‘-F"[t]"’ on the command line if you really do want to separate your fields with ‘t’s.

For example, let's use an awk program file called baud.awk that contains the pattern /300/ and the action ‘print $1’:

     /300/   { print $1 }

Let's also set FS to be the ‘-’ character and run the program on the file BBS-list. The following command prints a list of the names of the bulletin boards that operate at 300 baud and the first three digits of their phone numbers:

     $ awk -F- -f baud.awk BBS-list
     -| aardvark     555
     -| alpo
     -| barfly       555
     -| bites        555
     -| camelot      555
     -| core         555
     -| fooey        555
     -| foot         555
     -| macfoo       555
     -| sdace        555
     -| sabafoo      555

Note the second line of output. The second line in the original file looked like this:

     alpo-net     555-3412     2400/1200/300     A

The ‘-’ as part of the system's name was used as the field separator, instead of the ‘-’ in the phone number that was originally intended. This demonstrates why you have to be careful in choosing your field and record separators.

Perhaps the most common use of a single character as the field separator occurs when processing the Unix system password file. On many Unix systems, each user has a separate entry in the system password file, one line per user. The information in these lines is separated by colons. The first field is the user's login name and the second is the user's (encrypted or shadow) password. A password file entry might look like this:

     arnold:xyzzy:2076:10:Arnold Robbins:/home/arnold:/bin/bash

The following program searches the system password file and prints the entries for users who have no password:

     awk -F: '$2 == ""' /etc/passwd

Previous: Command Line Field Separator, Up: Field Separators

3.5.4 Field-Splitting Summary

It is important to remember that when you assign a string constant as the value of FS, it undergoes normal awk string processing. For example, with Unix awk and gawk, the assignment ‘FS = "\.."’ assigns the character string ".." to FS (the backslash is stripped). This creates a regexp meaning “fields are separated by occurrences of any two characters.” If instead you want fields to be separated by a literal period followed by any single character, use ‘FS = "\\.."’.

The following table summarizes how fields are split, based on the value of FS (‘==’ means “is equal to”):

FS == " "
Fields are separated by runs of whitespace. Leading and trailing whitespace are ignored. This is the default.
FS == any other single character
Fields are separated by each occurrence of the character. Multiple successive occurrences delimit empty fields, as do leading and trailing occurrences. The character can even be a regexp metacharacter; it does not need to be escaped.
FS == regexp
Fields are separated by occurrences of characters that match regexp. Leading and trailing matches of regexp delimit empty fields.
FS == ""
Each individual character in the record becomes a separate field. (This is a gawk extension; it is not specified by the POSIX standard.)

Advanced Notes: Changing FS Does Not Affect the Fields

According to the POSIX standard, awk is supposed to behave as if each record is split into fields at the time it is read. In particular, this means that if you change the value of FS after a record is read, the value of the fields (i.e., how they were split) should reflect the old value of FS, not the new one.

However, many implementations of awk do not work this way. Instead, they defer splitting the fields until a field is actually referenced. The fields are split using the current value of FS! (d.c.) This behavior can be difficult to diagnose. The following example illustrates the difference between the two methods. (The sed18 command prints just the first line of /etc/passwd.)

     sed 1q /etc/passwd | awk '{ FS = ":" ; print $1 }'

which usually prints:


on an incorrect implementation of awk, while gawk prints something like:


Advanced Notes: FS and IGNORECASE

The IGNORECASE variable (see User-modified) affects field splitting only when the value of FS is a regexp. It has no effect when FS is a single character, even if that character is a letter. Thus, in the following code:

     FS = "c"
     $0 = "aCa"
     print $1

The output is ‘aCa’. If you really want to split fields on an alphabetic character while ignoring case, use a regexp that will do it for you. E.g., ‘FS = "[c]"’. In this case, IGNORECASE will take effect.

Next: , Previous: Field Separators, Up: Reading Files

3.6 Reading Fixed-Width Data

NOTE: This section discusses an advanced feature of gawk. If you are a novice awk user, you might want to skip it on the first reading.

gawk version 2.13 introduced a facility for dealing with fixed-width fields with no distinctive field separator. For example, data of this nature arises in the input for old Fortran programs where numbers are run together, or in the output of programs that did not anticipate the use of their output as input for other programs.

An example of the latter is a table where all the columns are lined up by the use of a variable number of spaces and empty fields are just spaces. Clearly, awk's normal field splitting based on FS does not work well in this case. Although a portable awk program can use a series of substr calls on $0 (see String Functions), this is awkward and inefficient for a large number of fields.

The splitting of an input record into fixed-width fields is specified by assigning a string containing space-separated numbers to the built-in variable FIELDWIDTHS. Each number specifies the width of the field, including columns between fields. If you want to ignore the columns between fields, you can specify the width as a separate field that is subsequently ignored. It is a fatal error to supply a field width that is not a positive number. The following data is the output of the Unix w utility. It is useful to illustrate the use of FIELDWIDTHS:

      10:06pm  up 21 days, 14:04,  23 users
     User     tty       login  idle   JCPU   PCPU  what
     hzuo     ttyV0     8:58pm            9      5  vi p24.tex
     hzang    ttyV3     6:37pm    50                -csh
     eklye    ttyV5     9:53pm            7      1  em thes.tex
     dportein ttyV6     8:17pm  1:47                -csh
     gierd    ttyD3    10:00pm     1                elm
     dave     ttyD4     9:47pm            4      4  w
     brent    ttyp0    26Jun91  4:46  26:46   4:41  bash
     dave     ttyq4    26Jun9115days     46     46  wnewmail

The following program takes the above input, converts the idle time to number of seconds, and prints out the first two fields and the calculated idle time:

NOTE: This program uses a number of awk features that haven't been introduced yet.
     BEGIN  { FIELDWIDTHS = "9 6 10 6 7 7 35" }
     NR > 2 {
         idle = $4
         sub(/^  */, "", idle)   # strip leading spaces
         if (idle == "")
             idle = 0
         if (idle ~ /:/) {
             split(idle, t, ":")
             idle = t[1] * 60 + t[2]
         if (idle ~ /days/)
             idle *= 24 * 60 * 60
         print $1, $2, idle

Running the program on the data produces the following results:

     hzuo      ttyV0  0
     hzang     ttyV3  50
     eklye     ttyV5  0
     dportein  ttyV6  107
     gierd     ttyD3  1
     dave      ttyD4  0
     brent     ttyp0  286
     dave      ttyq4  1296000

Another (possibly more practical) example of fixed-width input data is the input from a deck of balloting cards. In some parts of the United States, voters mark their choices by punching holes in computer cards. These cards are then processed to count the votes for any particular candidate or on any particular issue. Because a voter may choose not to vote on some issue, any column on the card may be empty. An awk program for processing such data could use the FIELDWIDTHS feature to simplify reading the data. (Of course, getting gawk to run on a system with card readers is another story!)

Assigning a value to FS causes gawk to use FS for field splitting again. Use ‘FS = FS’ to make this happen, without having to know the current value of FS. In order to tell which kind of field splitting is in effect, use PROCINFO["FS"] (see Auto-set). The value is "FS" if regular field splitting is being used, or it is "FIELDWIDTHS" if fixed-width field splitting is being used:

     if (PROCINFO["FS"] == "FS")
         regular field splitting ...
         fixed-width field splitting ...

This information is useful when writing a function that needs to temporarily change FS or FIELDWIDTHS, read some records, and then restore the original settings (see Passwd Functions, for an example of such a function).

Next: , Previous: Constant Size, Up: Reading Files

3.7 Multiple-Line Records

In some databases, a single line cannot conveniently hold all the information in one entry. In such cases, you can use multiline records. The first step in doing this is to choose your data format.

One technique is to use an unusual character or string to separate records. For example, you could use the formfeed character (written ‘\f’ in awk, as in C) to separate them, making each record a page of the file. To do this, just set the variable RS to "\f" (a string containing the formfeed character). Any other character could equally well be used, as long as it won't be part of the data in a record.

Another technique is to have blank lines separate records. By a special dispensation, an empty string as the value of RS indicates that records are separated by one or more blank lines. When RS is set to the empty string, each record always ends at the first blank line encountered. The next record doesn't start until the first nonblank line that follows. No matter how many blank lines appear in a row, they all act as one record separator. (Blank lines must be completely empty; lines that contain only whitespace do not count.)

You can achieve the same effect as ‘RS = ""’ by assigning the string "\n\n+" to RS. This regexp matches the newline at the end of the record and one or more blank lines after the record. In addition, a regular expression always matches the longest possible sequence when there is a choice (see Leftmost Longest). So the next record doesn't start until the first nonblank line that follows—no matter how many blank lines appear in a row, they are considered one record separator.

There is an important difference between ‘RS = ""’ and ‘RS = "\n\n+"’. In the first case, leading newlines in the input data file are ignored, and if a file ends without extra blank lines after the last record, the final newline is removed from the record. In the second case, this special processing is not done. (d.c.)

Now that the input is separated into records, the second step is to separate the fields in the record. One way to do this is to divide each of the lines into fields in the normal manner. This happens by default as the result of a special feature. When RS is set to the empty string, and FS is set to a single character, the newline character always acts as a field separator. This is in addition to whatever field separations result from FS.19

The original motivation for this special exception was probably to provide useful behavior in the default case (i.e., FS is equal to " "). This feature can be a problem if you really don't want the newline character to separate fields, because there is no way to prevent it. However, you can work around this by using the split function to break up the record manually (see String Functions). If you have a single character field separator, you can work around the special feature in a different way, by making FS into a regexp for that single character. For example, if the field separator is a percent character, instead of ‘FS = "%"’, use ‘FS = "[%]"’.

Another way to separate fields is to put each field on a separate line: to do this, just set the variable FS to the string "\n". (This single character separator matches a single newline.) A practical example of a data file organized this way might be a mailing list, where each entry is separated by blank lines. Consider a mailing list in a file named addresses, which looks like this:

     Jane Doe
     123 Main Street
     Anywhere, SE 12345-6789
     John Smith
     456 Tree-lined Avenue
     Smallville, MW 98765-4321

A simple program to process this file is as follows:

     # addrs.awk --- simple mailing list program
     # Records are separated by blank lines.
     # Each line is one field.
     BEGIN { RS = "" ; FS = "\n" }
           print "Name is:", $1
           print "Address is:", $2
           print "City and State are:", $3
           print ""

Running the program produces the following output:

     $ awk -f addrs.awk addresses
     -| Name is: Jane Doe
     -| Address is: 123 Main Street
     -| City and State are: Anywhere, SE 12345-6789
     -| Name is: John Smith
     -| Address is: 456 Tree-lined Avenue
     -| City and State are: Smallville, MW 98765-4321

See Labels Program, for a more realistic program that deals with address lists. The following table summarizes how records are split, based on the value of RS:

RS == "\n"
Records are separated by the newline character (‘\n’). In effect, every line in the data file is a separate record, including blank lines. This is the default.
RS == any single character
Records are separated by each occurrence of the character. Multiple successive occurrences delimit empty records.
RS == ""
Records are separated by runs of blank lines. When FS is a single character, then the newline character always serves as a field separator, in addition to whatever value FS may have. Leading and trailing newlines in a file are ignored.
RS == regexp
Records are separated by occurrences of characters that match regexp. Leading and trailing matches of regexp delimit empty records. (This is a gawk extension; it is not specified by the POSIX standard.)

In all cases, gawk sets RT to the input text that matched the value specified by RS.

Previous: Multiple Line, Up: Reading Files

3.8 Explicit Input with getline

So far we have been getting our input data from awk's main input stream—either the standard input (usually your terminal, sometimes the output from another program) or from the files specified on the command line. The awk language has a special built-in command called getline that can be used to read input under your explicit control.

The getline command is used in several different ways and should not be used by beginners. The examples that follow the explanation of the getline command include material that has not been covered yet. Therefore, come back and study the getline command after you have reviewed the rest of this Web page and have a good knowledge of how awk works.

The getline command returns one if it finds a record and zero if it encounters the end of the file. If there is some error in getting a record, such as a file that cannot be opened, then getline returns −1. In this case, gawk sets the variable ERRNO to a string describing the error that occurred.

In the following examples, command stands for a string value that represents a shell command.

Next: , Up: Getline

3.8.1 Using getline with No Arguments

The getline command can be used without arguments to read input from the current input file. All it does in this case is read the next input record and split it up into fields. This is useful if you've finished processing the current record, but want to do some special processing on the next record right now. For example:

          if ((t = index($0, "/*")) != 0) {
               # value of `tmp' will be "" if t is 1
               tmp = substr($0, 1, t - 1)
               u = index(substr($0, t + 2), "*/")
               while (u == 0) {
                    if (getline <= 0) {
                         m = "unexpected EOF or error"
                         m = (m ": " ERRNO)
                         print m > "/dev/stderr"
                    u = index($0, "*/")
               # substr expression will be "" if */
               # occurred at end of line
               $0 = tmp substr($0, u + 2)
          print $0

This awk program deletes C-style comments (‘/* ... */’) from the input. By replacing the ‘print $0’ with other statements, you could perform more complicated processing on the decommented input, such as searching for matches of a regular expression. (This program has a subtle problem—it does not work if one comment ends and another begins on the same line.)

This form of the getline command sets NF, NR, FNR, and the value of $0.

NOTE: The new value of $0 is used to test the patterns of any subsequent rules. The original value of $0 that triggered the rule that executed getline is lost. By contrast, the next statement reads a new record but immediately begins processing it normally, starting with the first rule in the program. See Next Statement.

Next: , Previous: Plain Getline, Up: Getline

3.8.2 Using getline into a Variable

You can use ‘getline var’ to read the next record from awk's input into the variable var. No other processing is done. For example, suppose the next line is a comment or a special string, and you want to read it without triggering any rules. This form of getline allows you to read that line and store it in a variable so that the main read-a-line-and-check-each-rule loop of awk never sees it. The following example swaps every two lines of input:

          if ((getline tmp) > 0) {
               print tmp
               print $0
          } else
               print $0

It takes the following list:


and produces these results:


The getline command used in this way sets only the variables NR and FNR (and of course, var). The record is not split into fields, so the values of the fields (including $0) and the value of NF do not change.

Next: , Previous: Getline/Variable, Up: Getline

3.8.3 Using getline from a File

Use ‘getline < file’ to read the next record from file. Here file is a string-valued expression that specifies the file name. ‘< file’ is called a redirection because it directs input to come from a different place. For example, the following program reads its input record from the file secondary.input when it encounters a first field with a value equal to 10 in the current input file:

         if ($1 == 10) {
              getline < "secondary.input"
         } else

Because the main input stream is not used, the values of NR and FNR are not changed. However, the record it reads is split into fields in the normal manner, so the values of $0 and the other fields are changed, resulting in a new value of NF.

According to POSIX, ‘getline < expression’ is ambiguous if expression contains unparenthesized operators other than ‘$’; for example, ‘getline < dir "/" file’ is ambiguous because the concatenation operator is not parenthesized. You should write it as ‘getline < (dir "/" file)’ if you want your program to be portable to other awk implementations.

Next: , Previous: Getline/File, Up: Getline

3.8.4 Using getline into a Variable from a File

Use ‘getline var < file’ to read input from the file file, and put it in the variable var. As above, file is a string-valued expression that specifies the file from which to read.

In this version of getline, none of the built-in variables are changed and the record is not split into fields. The only variable changed is var. For example, the following program copies all the input files to the output, except for records that say ‘@include filename. Such a record is replaced by the contents of the file filename:

          if (NF == 2 && $1 == "@include") {
               while ((getline line < $2) > 0)
                    print line
          } else

Note here how the name of the extra input file is not built into the program; it is taken directly from the data, specifically from the second field on the ‘@include’ line.

The close function is called to ensure that if two identical ‘@include’ lines appear in the input, the entire specified file is included twice. See Close Files And Pipes.

One deficiency of this program is that it does not process nested ‘@include’ statements (i.e., ‘@include’ statements in included files) the way a true macro preprocessor would. See Igawk Program, for a program that does handle nested ‘@include’ statements.

Next: , Previous: Getline/Variable/File, Up: Getline

3.8.5 Using getline from a Pipe

The output of a command can also be piped into getline, using ‘command | getline’. In this case, the string command is run as a shell command and its output is piped into awk to be used as input. This form of getline reads one record at a time from the pipe. For example, the following program copies its input to its output, except for lines that begin with ‘@execute’, which are replaced by the output produced by running the rest of the line as a shell command:

          if ($1 == "@execute") {
               tmp = substr($0, 10)
               while ((tmp | getline) > 0)
          } else

The close function is called to ensure that if two identical ‘@execute’ lines appear in the input, the command is run for each one. See Close Files And Pipes. Given the input:

     @execute who

the program might produce:

     arnold     ttyv0   Jul 13 14:22
     miriam     ttyp0   Jul 13 14:23     (murphy:0)
     bill       ttyp1   Jul 13 14:23     (murphy:0)

Notice that this program ran the command who and printed the previous result. (If you try this program yourself, you will of course get different results, depending upon who is logged in on your system.)

This variation of getline splits the record into fields, sets the value of NF, and recomputes the value of $0. The values of NR and FNR are not changed.

According to POSIX, ‘expression | getline’ is ambiguous if expression contains unparenthesized operators other than ‘$’—for example, ‘"echo " "date" | getline’ is ambiguous because the concatenation operator is not parenthesized. You should write it as ‘("echo " "date") | getline’ if you want your program to be portable to other awk implementations.

NOTE: Unfortunately, gawk has not been consistent in its treatment of a construct like ‘"echo " "date" | getline’. Up to and including version 3.1.1 of gawk, it was treated as ‘("echo " "date") | getline’. (This how Unix awk behaves.) From 3.1.2 through 3.1.5, it was treated as ‘"echo " ("date" | getline)’. (This is how mawk behaves.) Starting with version 3.1.6, the earlier behavior was reinstated. In short, always use explicit parentheses, and then you won't have to worry.

Next: , Previous: Getline/Pipe, Up: Getline

3.8.6 Using getline into a Variable from a Pipe

When you use ‘command | getline var’, the output of command is sent through a pipe to getline and into the variable var. For example, the following program reads the current date and time into the variable current_time, using the date utility, and then prints it:

     BEGIN {
          "date" | getline current_time
          print "Report printed on " current_time

In this version of getline, none of the built-in variables are changed and the record is not split into fields.

Next: , Previous: Getline/Variable/Pipe, Up: Getline

3.8.7 Using getline from a Coprocess

Input into getline from a pipe is a one-way operation. The command that is started with ‘command | getline’ only sends data to your awk program.

On occasion, you might want to send data to another program for processing and then read the results back. gawk allows you to start a coprocess, with which two-way communications are possible. This is done with the ‘|&’ operator. Typically, you write data to the coprocess first and then read results back, as shown in the following:

     print "some query" |& "db_server"
     "db_server" |& getline

which sends a query to db_server and then reads the results.

The values of NR and FNR are not changed, because the main input stream is not used. However, the record is split into fields in the normal manner, thus changing the values of $0, of the other fields, and of NF.

Coprocesses are an advanced feature. They are discussed here only because this is the section on getline. See Two-way I/O, where coprocesses are discussed in more detail.

Next: , Previous: Getline/Coprocess, Up: Getline

3.8.8 Using getline into a Variable from a Coprocess

When you use ‘command |& getline var’, the output from the coprocess command is sent through a two-way pipe to getline and into the variable var.

In this version of getline, none of the built-in variables are changed and the record is not split into fields. The only variable changed is var.

Next: , Previous: Getline/Variable/Coprocess, Up: Getline

3.8.9 Points to Remember About getline

Here are some miscellaneous points about getline that you should bear in mind:

Previous: Getline Notes, Up: Getline

3.8.10 Summary of getline Variants

table-getline-variants summarizes the eight variants of getline, listing which built-in variables are set by each one.

Variant Effect
getline Sets $0, NF, FNR, and NR
getline var Sets var, FNR, and NR
getline < file Sets $0 and NF
getline var < file Sets var
command | getline Sets $0 and NF
command | getline var Sets var
command |& getline Sets $0 and NF. This is a gawk extension
command |& getline var Sets var. This is a gawk extension

Table 3.1: getline Variants and What They Set

Next: , Previous: Reading Files, Up: Top

4 Printing Output

One of the most common programming actions is to print, or output, some or all of the input. Use the print statement for simple output, and the printf statement for fancier formatting. The print statement is not limited when computing which values to print. However, with two exceptions, you cannot specify how to print them—how many columns, whether to use exponential notation or not, and so on. (For the exceptions, see Output Separators, and OFMT.) For printing with specifications, you need the printf statement (see Printf).

Besides basic and formatted printing, this chapter also covers I/O redirections to files and pipes, introduces the special file names that gawk processes internally, and discusses the close built-in function.

Next: , Up: Printing

4.1 The print Statement

The print statement is used to produce output with simple, standardized formatting. Specify only the strings or numbers to print, in a list separated by commas. They are output, separated by single spaces, followed by a newline. The statement looks like this:

     print item1, item2, ...

The entire list of items may be optionally enclosed in parentheses. The parentheses are necessary if any of the item expressions uses the ‘>’ relational operator; otherwise it could be confused with a redirection (see Redirection).

The items to print can be constant strings or numbers, fields of the current record (such as $1), variables, or any awk expression. Numeric values are converted to strings and then printed.

The simple statement ‘print’ with no items is equivalent to ‘print $0’: it prints the entire current record. To print a blank line, use ‘print ""’, where "" is the empty string. To print a fixed piece of text, use a string constant, such as "Don't Panic", as one item. If you forget to use the double-quote characters, your text is taken as an awk expression, and you will probably get an error. Keep in mind that a space is printed between any two items.

Next: , Previous: Print, Up: Printing

4.2 Examples of print Statements

Each print statement makes at least one line of output. However, it isn't limited to only one line. If an item value is a string that contains a newline, the newline is output along with the rest of the string. A single print statement can make any number of lines this way.

The following is an example of printing a string that contains embedded newlines (the ‘\n’ is an escape sequence, used to represent the newline character; see Escape Sequences):

     $ awk 'BEGIN { print "line one\nline two\nline three" }'
     -| line one
     -| line two
     -| line three

The next example, which is run on the inventory-shipped file, prints the first two fields of each input record, with a space between them:

     $ awk '{ print $1, $2 }' inventory-shipped
     -| Jan 13
     -| Feb 15
     -| Mar 15

A common mistake in using the print statement is to omit the comma between two items. This often has the effect of making the items run together in the output, with no space. The reason for this is that juxtaposing two string expressions in awk means to concatenate them. Here is the same program, without the comma:

     $ awk '{ print $1 $2 }' inventory-shipped
     -| Jan13
     -| Feb15
     -| Mar15

To someone unfamiliar with the inventory-shipped file, neither example's output makes much sense. A heading line at the beginning would make it clearer. Let's add some headings to our table of months ($1) and green crates shipped ($2). We do this using the BEGIN pattern (see BEGIN/END) so that the headings are only printed once:

     awk 'BEGIN {  print "Month Crates"
                   print "----- ------" }
                {  print $1, $2 }' inventory-shipped

When run, the program prints the following:

     Month Crates
     ----- ------
     Jan 13
     Feb 15
     Mar 15

The only problem, however, is that the headings and the table data don't line up! We can fix this by printing some spaces between the two fields:

     awk 'BEGIN { print "Month Crates"
                  print "----- ------" }
                { print $1, "     ", $2 }' inventory-shipped

Lining up columns this way can get pretty complicated when there are many columns to fix. Counting spaces for two or three columns is simple, but any more than this can take up a lot of time. This is why the printf statement was created (see Printf); one of its specialties is lining up columns of data.

NOTE: You can continue either a print or printf statement simply by putting a newline after any comma (see Statements/Lines).

Next: , Previous: Print Examples, Up: Printing

4.3 Output Separators

As mentioned previously, a print statement contains a list of items separated by commas. In the output, the items are normally separated by single spaces. However, this doesn't need to be the case; a single space is only the default. Any string of characters may be used as the output field separator by setting the built-in variable OFS. The initial value of this variable is the string " "—that is, a single space.

The output from an entire print statement is called an output record. Each print statement outputs one output record, and then outputs a string called the output record separator (or ORS). The initial value of ORS is the string "\n"; i.e., a newline character. Thus, each print statement normally makes a separate line.

In order to change how output fields and records are separated, assign new values to the variables OFS and ORS. The usual place to do this is in the BEGIN rule (see BEGIN/END), so that it happens before any input is processed. It can also be done with assignments on the command line, before the names of the input files, or using the -v command-line option (see Options). The following example prints the first and second fields of each input record, separated by a semicolon, with a blank line added after each newline:

     $ awk 'BEGIN { OFS = ";"; ORS = "\n\n" }
     >            { print $1, $2 }' BBS-list
     -| aardvark;555-5553
     -| alpo-net;555-3412
     -| barfly;555-7685

If the value of ORS does not contain a newline, the program's output is run together on a single line.

Next: , Previous: Output Separators, Up: Printing

4.4 Controlling Numeric Output with print

When the print statement is used to print numeric values, awk internally converts the number to a string of characters and prints that string. awk uses the sprintf function to do this conversion (see String Functions). For now, it suffices to say that the sprintf function accepts a format specification that tells it how to format numbers (or strings), and that there are a number of different ways in which numbers can be formatted. The different format specifications are discussed more fully in Control Letters.

The built-in variable OFMT contains the default format specification that print uses with sprintf when it wants to convert a number to a string for printing. The default value of OFMT is "%.6g". The way print prints numbers can be changed by supplying different format specifications as the value of OFMT, as shown in the following example:

     $ awk 'BEGIN {
     >   OFMT = "%.0f"  # print numbers as integers (rounds)
     >   print 17.23, 17.54 }'
     -| 17 18

According to the POSIX standard, awk's behavior is undefined if OFMT contains anything but a floating-point conversion specification. (d.c.)

Next: , Previous: OFMT, Up: Printing

4.5 Using printf Statements for Fancier Printing

For more precise control over the output format than what is normally provided by print, use printf. printf can be used to specify the width to use for each item, as well as various formatting choices for numbers (such as what output base to use, whether to print an exponent, whether to print a sign, and how many digits to print after the decimal point). This is done by supplying a string, called the format string, that controls how and where to print the other arguments.

Next: , Up: Printf

4.5.1 Introduction to the printf Statement

A simple printf statement looks like this:

     printf format, item1, item2, ...

The entire list of arguments may optionally be enclosed in parentheses. The parentheses are necessary if any of the item expressions use the ‘>’ relational operator; otherwise, it can be confused with a redirection (see Redirection).

The difference between printf and print is the format argument. This is an expression whose value is taken as a string; it specifies how to output each of the other arguments. It is called the format string.

The format string is very similar to that in the ISO C library function printf. Most of format is text to output verbatim. Scattered among this text are format specifiers—one per item. Each format specifier says to output the next item in the argument list at that place in the format.

The printf statement does not automatically append a newline to its output. It outputs only what the format string specifies. So if a newline is needed, you must include one in the format string. The output separator variables OFS and ORS have no effect on printf statements. For example:

     $ awk 'BEGIN {
     >    ORS = "\nOUCH!\n"; OFS = "+"
     >    msg = "Dont Panic!"
     >    printf "%s\n", msg
     > }'
     -| Dont Panic!

Here, neither the ‘+’ nor the ‘OUCH’ appear when the message is printed.

Next: , Previous: Basic Printf, Up: Printf

4.5.2 Format-Control Letters

A format specifier starts with the character ‘%’ and ends with a format-control letter—it tells the printf statement how to output one item. The format-control letter specifies what kind of value to print. The rest of the format specifier is made up of optional modifiers that control how to print the value, such as the field width. Here is a list of the format-control letters:

This prints a number as an ASCII character; thus, ‘printf "%c", 65’ outputs the letter ‘A’. (The output for a string value is the first character of the string.)
%d, %i
These are equivalent; they both print a decimal integer. (The ‘%i’ specification is for compatibility with ISO C.)
%e, %E
These print a number in scientific (exponential) notation; for example:
          printf "%4.3e\n", 1950

prints ‘1.950e+03’, with a total of four significant figures, three of which follow the decimal point. (The ‘4.3’ represents two modifiers, discussed in the next subsection.) ‘%E’ uses ‘E’ instead of ‘e’ in the output.

This prints a number in floating-point notation. For example:
          printf "%4.3f", 1950

prints ‘1950.000’, with a total of four significant figures, three of which follow the decimal point. (The ‘4.3’ represents two modifiers, discussed in the next subsection.)

On systems supporting IEEE 754 floating point format, values representing negative infinity are formatted as ‘-inf’ or ‘-infinity’, and positive infinity as ‘inf’ and ‘infinity’. The special “not a number” value formats as ‘-nan’ or ‘nan’.

Like %f but the infinity and “not a number” values are spelled using uppercase letters.

The %F format is a POSIX extension to ISO C; not all systems support it. On those that don't, gawk uses %f instead.

%g, %G
These print a number in either scientific notation or in floating-point notation, whichever uses fewer characters; if the result is printed in scientific notation, ‘%G’ uses ‘E’ instead of ‘e’.
This prints an unsigned octal integer.
This prints a string.
This prints an unsigned decimal integer. (This format is of marginal use, because all numbers in awk are floating-point; it is provided primarily for compatibility with C.)
%x, %X
These print an unsigned hexadecimal integer; ‘%X’ uses the letters ‘A’ through ‘F’ instead of ‘a’ through ‘f’.
This isn't a format-control letter, but it does have meaning—the sequence ‘%%’ outputs one ‘%’; it does not consume an argument and it ignores any modifiers.

NOTE: When using the integer format-control letters for values that are outside the range of the widest C integer type, gawk switches to the ‘%g’ format specifier. If --lint is provided on the command line (see Options), gawk warns about this. Other versions of awk may print invalid values or do something else entirely. (d.c.)

Next: , Previous: Control Letters, Up: Printf

4.5.3 Modifiers for printf Formats

A format specification can also include modifiers that can control how much of the item's value is printed, as well as how much space it gets. The modifiers come between the ‘%’ and the format-control letter. We will use the bullet symbol “•” in the following examples to represent spaces in the output. Here are the possible modifiers, in the order in which they may appear:

An integer constant followed by a ‘$’ is a positional specifier. Normally, format specifications are applied to arguments in the order given in the format string. With a positional specifier, the format specification is applied to a specific argument, instead of what would be the next argument in the list. Positional specifiers begin counting with one. Thus:
          printf "%s %s\n", "don't", "panic"
          printf "%2$s %1$s\n", "panic", "don't"

prints the famous friendly message twice.

At first glance, this feature doesn't seem to be of much use. It is in fact a gawk extension, intended for use in translating messages at runtime. See Printf Ordering, which describes how and why to use positional specifiers. For now, we will not use them.

The minus sign, used before the width modifier (see later on in this table), says to left-justify the argument within its specified width. Normally, the argument is printed right-justified in the specified width. Thus:
          printf "%-4s", "foo"

prints ‘foo•’.

For numeric conversions, prefix positive values with a space and negative values with a minus sign.
The plus sign, used before the width modifier (see later on in this table), says to always supply a sign for numeric conversions, even if the data to format is positive. The ‘+’ overrides the space modifier.
Use an “alternate form” for certain control letters. For ‘%o’, supply a leading zero. For ‘%x’ and ‘%X’, supply a leading ‘0x’ or ‘0X’ for a nonzero result. For ‘%e’, ‘%E’, and ‘%f’, the result always contains a decimal point. For ‘%g’ and ‘%G’, trailing zeros are not removed from the result.

A leading ‘0’ (zero) acts as a flag that indicates that output should be padded with zeros instead of spaces. This applies even to non-numeric output formats. (d.c.) This flag only has an effect when the field width is wider than the value to print.
A single quote or apostrophe character is a POSIX extension to ISO C. It indicates that the integer part of a floating point value, or the entire part of an integer decimal value, should have a thousands-separator character in it. This only works in locales that support such characters. For example:
          $ cat thousands.awk          Show source program
          -| BEGIN { printf "%'d\n", 1234567 }
          $ LC_ALL=C gawk -f thousands.awk
          -| 1234567                   Results in "C" locale
          $ LC_ALL=en_US.UTF-8 gawk -f thousands.awk
          -| 1,234,567                 Results in US English UTF locale

For more information about locales and internationalization issues, see Locales.

NOTE: The ‘'’ flag is a nice feature, but its use complicates things: it becomes difficult to use it in command-line programs. For information on appropriate quoting tricks, see Quoting.

This is a number specifying the desired minimum width of a field. Inserting any number between the ‘%’ sign and the format-control character forces the field to expand to this width. The default way to do this is to pad with spaces on the left. For example:
          printf "%4s", "foo"

prints ‘•foo’.

The value of width is a minimum width, not a maximum. If the item value requires more than width characters, it can be as wide as necessary. Thus, the following:

          printf "%4s", "foobar"

prints ‘foobar’.

Preceding the width with a minus sign causes the output to be padded with spaces on the right, instead of on the left.

A period followed by an integer constant specifies the precision to use when printing. The meaning of the precision varies by control letter:
%e, %E, %f
Number of digits to the right of the decimal point.
%g, %G
Maximum number of significant digits.
%d, %i, %o, %u, %x, %X
Minimum number of digits to print.
Maximum number of characters from the string that should print.

Thus, the following:

          printf "%.4s", "foobar"

prints ‘foob’.

The C library printf's dynamic width and prec capability (for example, "%*.*s") is supported. Instead of supplying explicit width and/or prec values in the format string, they are passed in the argument list. For example:

     w = 5
     p = 3
     s = "abcdefg"
     printf "%*.*s\n", w, p, s

is exactly equivalent to:

     s = "abcdefg"
     printf "%5.3s\n", s

Both programs output ‘••abc’. Earlier versions of awk did not support this capability. If you must use such a version, you may simulate this feature by using concatenation to build up the format string, like so:

     w = 5
     p = 3
     s = "abcdefg"
     printf "%" w "." p "s\n", s

This is not particularly easy to read but it does work.

C programmers may be used to supplying additional ‘l’, ‘L’, and ‘h’ modifiers in printf format strings. These are not valid in awk. Most awk implementations silently ignore these modifiers. If --lint is provided on the command line (see Options), gawk warns about their use. If --posix is supplied, their use is a fatal error.

Previous: Format Modifiers, Up: Printf

4.5.4 Examples Using printf

The following is a simple example of how to use printf to make an aligned table:

     awk '{ printf "%-10s %s\n", $1, $2 }' BBS-list

This command prints the names of the bulletin boards ($1) in the file BBS-list as a string of 10 characters that are left-justified. It also prints the phone numbers ($2) next on the line. This produces an aligned two-column table of names and phone numbers, as shown here:

     $ awk '{ printf "%-10s %s\n", $1, $2 }' BBS-list
     -| aardvark   555-5553
     -| alpo-net   555-3412
     -| barfly     555-7685
     -| bites      555-1675
     -| camelot    555-0542
     -| core       555-2912
     -| fooey      555-1234
     -| foot       555-6699
     -| macfoo     555-6480
     -| sdace      555-3430
     -| sabafoo    555-2127

In this case, the phone numbers had to be printed as strings because the numbers are separated by a dash. Printing the phone numbers as numbers would have produced just the first three digits: ‘555’. This would have been pretty confusing.

It wasn't necessary to specify a width for the phone numbers because they are last on their lines. They don't need to have spaces after them.

The table could be made to look even nicer by adding headings to the tops of the columns. This is done using the BEGIN pattern (see BEGIN/END) so that the headers are only printed once, at the beginning of the awk program:

     awk 'BEGIN { print "Name      Number"
                  print "----      ------" }
          { printf "%-10s %s\n", $1, $2 }' BBS-list

The above example mixed print and printf statements in the same program. Using just printf statements can produce the same results:

     awk 'BEGIN { printf "%-10s %s\n", "Name", "Number"
                  printf "%-10s %s\n", "----", "------" }
          { printf "%-10s %s\n", $1, $2 }' BBS-list

Printing each column heading with the same format specification used for the column elements ensures that the headings are aligned just like the columns.

The fact that the same format specification is used three times can be emphasized by storing it in a variable, like this:

     awk 'BEGIN { format = "%-10s %s\n"
                  printf format, "Name", "Number"
                  printf format, "----", "------" }
          { printf format, $1, $2 }' BBS-list

At this point, it would be a worthwhile exercise to use the printf statement to line up the headings and table data for the inventory-shipped example that was covered earlier in the section on the print statement (see Print).

Next: , Previous: Printf, Up: Printing

4.6 Redirecting Output of print and printf

So far, the output from print and printf has gone to the standard output, usually the terminal. Both print and printf can also send their output to other places. This is called redirection.

A redirection appears after the print or printf statement. Redirections in awk are written just like redirections in shell commands, except that they are written inside the awk program.

There are four forms of output redirection: output to a file, output appended to a file, output through a pipe to another command, and output to a coprocess. They are all shown for the print statement, but they work identically for printf:

print items > output-file
This type of redirection prints the items into the output file named output-file. The file name output-file can be any expression. Its value is changed to a string and then used as a file name (see Expressions).

When this type of redirection is used, the output-file is erased before the first output is written to it. Subsequent writes to the same output-file do not erase output-file, but append to it. (This is different from how you use redirections in shell scripts.) If output-file does not exist, it is created. For example, here is how an awk program can write a list of BBS names to one file named name-list, and a list of phone numbers to another file named phone-list:

          $ awk '{ print $2 > "phone-list"
          >        print $1 > "name-list" }' BBS-list
          $ cat phone-list
          -| 555-5553
          -| 555-3412
          $ cat name-list
          -| aardvark
          -| alpo-net

Each output file contains one name or number per line.

print items >> output-file
This type of redirection prints the items into the pre-existing output file named output-file. The difference between this and the single-‘>’ redirection is that the old contents (if any) of output-file are not erased. Instead, the awk output is appended to the file. If output-file does not exist, then it is created.

print items | command
It is also possible to send output to another program through a pipe instead of into a file. This type of redirection opens a pipe to command, and writes the values of items through this pipe to another process created to execute command.

The redirection argument command is actually an awk expression. Its value is converted to a string whose contents give the shell command to be run. For example, the following produces two files, one unsorted list of BBS names, and one list sorted in reverse alphabetical order:

          awk '{ print $1 > "names.unsorted"
                 command = "sort -r > names.sorted"
                 print $1 | command }' BBS-list

The unsorted list is written with an ordinary redirection, while the sorted list is written by piping through the sort utility.

The next example uses redirection to mail a message to the mailing list ‘bug-system’. This might be useful when trouble is encountered in an awk script run periodically for system maintenance:

          report = "mail bug-system"
          print "Awk script failed:", $0 | report
          m = ("at record number " FNR " of " FILENAME)
          print m | report

The message is built using string concatenation and saved in the variable m. It's then sent down the pipeline to the mail program. (The parentheses group the items to concatenate—see Concatenation.)

The close function is called here because it's a good idea to close the pipe as soon as all the intended output has been sent to it. See Close Files And Pipes, for more information.

This example also illustrates the use of a variable to represent a file or command—it is not necessary to always use a string constant. Using a variable is generally a good idea, because (if you mean to refer to that same file or command) awk requires that the string value be spelled identically every time.

print items |& command
This type of redirection prints the items to the input of command. The difference between this and the single-‘|’ redirection is that the output from command can be read with getline. Thus command is a coprocess, which works together with, but subsidiary to, the awk program.

This feature is a gawk extension, and is not available in POSIX awk. See Getline/Coprocess, for a brief discussion. See Two-way I/O, for a more complete discussion.

Redirecting output using ‘>’, ‘>>’, ‘|’, or ‘|&’ asks the system to open a file, pipe, or coprocess only if the particular file or command you specify has not already been written to by your program or if it has been closed since it was last written to.

It is a common error to use ‘>’ redirection for the first print to a file, and then to use ‘>>’ for subsequent output:

     # clear the file
     print "Don't panic" > "guide.txt"
     # append
     print "Avoid improbability generators" >> "guide.txt"

This is indeed how redirections must be used from the shell. But in awk, it isn't necessary. In this kind of case, a program should use ‘>’ for all the print statements, since the output file is only opened once. (It happens that if you mix ‘>’ and ‘>>’ that output is produced in the expected order. However, mixing the operators for the same file is definitely poor style, and is confusing to readers of your program.)

As mentioned earlier (see Getline Notes), many Many awk implementations limit the number of pipelines that an awk program may have open to just one! In gawk, there is no such limit. gawk allows a program to open as many pipelines as the underlying operating system permits.

Advanced Notes: Piping into sh

A particularly powerful way to use redirection is to build command lines and pipe them into the shell, sh. For example, suppose you have a list of files brought over from a system where all the file names are stored in uppercase, and you wish to rename them to have names in all lowercase. The following program is both simple and efficient:

     { printf("mv %s %s\n", $0, tolower($0)) | "sh" }
     END { close("sh") }

The tolower function returns its argument string with all uppercase characters converted to lowercase (see String Functions). The program builds up a list of command lines, using the mv utility to rename the files. It then sends the list to the shell for execution.

Next: , Previous: Redirection, Up: Printing

4.7 Special File Names in gawk

gawk provides a number of special file names that it interprets internally. These file names provide access to standard file descriptors, process-related information, and TCP/IP networking.

Next: , Up: Special Files

4.7.1 Special Files for Standard Descriptors

Running programs conventionally have three input and output streams already available to them for reading and writing. These are known as the standard input, standard output, and standard error output. These streams are, by default, connected to your terminal, but they are often redirected with the shell, via the ‘<’, ‘<<’, ‘>’, ‘>>’, ‘>&’, and ‘|’ operators. Standard error is typically used for writing error messages; the reason there are two separate streams, standard output and standard error, is so that they can be redirected separately.

In other implementations of awk, the only way to write an error message to standard error in an awk program is as follows:

     print "Serious error detected!" | "cat 1>&2"

This works by opening a pipeline to a shell command that can access the standard error stream that it inherits from the awk process. This is far from elegant, and it is also inefficient, because it requires a separate process. So people writing awk programs often don't do this. Instead, they send the error messages to the terminal, like this:

     print "Serious error detected!" > "/dev/tty"

This usually has the same effect but not always: although the standard error stream is usually the terminal, it can be redirected; when that happens, writing to the terminal is not correct. In fact, if awk is run from a background job, it may not have a terminal at all. Then opening /dev/tty fails.

gawk provides special file names for accessing the three standard streams, as well as any other inherited open files. If the file name matches one of these special names when gawk redirects input or output, then it directly uses the stream that the file name stands for. These special file names work for all operating systems that gawk has been ported to, not just those that are POSIX-compliant:

The standard input (file descriptor 0).
The standard output (file descriptor 1).
The standard error output (file descriptor 2).
The file associated with file descriptor N. Such a file must be opened by the program initiating the awk execution (typically the shell). Unless special pains are taken in the shell from which gawk is invoked, only descriptors 0, 1, and 2 are available.

The file names /dev/stdin, /dev/stdout, and /dev/stderr are aliases for /dev/fd/0, /dev/fd/1, and /dev/fd/2, respectively. However, they are more self-explanatory. The proper way to write an error message in a gawk program is to use /dev/stderr, like this:

     print "Serious error detected!" > "/dev/stderr"

Note the use of quotes around the file name. Like any other redirection, the value must be a string. It is a common error to omit the quotes, which leads to confusing results.

Next: , Previous: Special FD, Up: Special Files

4.7.2 Special Files for Process-Related Information

gawk also provides special file names that give access to information about the running gawk process. Each of these “files” provides a single record of information. To read them more than once, they must first be closed with the close function (see Close Files And Pipes). The file names are:

Reading this file returns the process ID of the current process, in decimal form, terminated with a newline.
Reading this file returns the parent process ID of the current process, in decimal form, terminated with a newline.
Reading this file returns the process group ID of the current process, in decimal form, terminated with a newline.
Reading this file returns a single record terminated with a newline. The fields are separated with spaces. The fields represent the following information:
The return value of the getuid system call (the real user ID number).
The return value of the geteuid system call (the effective user ID number).
The return value of the getgid system call (the real group ID number).
The return value of the getegid system call (the effective group ID number).

If there are any additional fields, they are the group IDs returned by the getgroups system call. (Multiple groups may not be supported on all systems.)

These special file names may be used on the command line as data files, as well as for I/O redirections within an awk program. They may not be used as source files with the -f option.

NOTE: The special files that provide process-related information are now considered obsolete and will disappear entirely in the next release of gawk. gawk prints a warning message every time you use one of these files. To obtain process-related information, use the PROCINFO array. See Auto-set.

Next: , Previous: Special Process, Up: Special Files

4.7.3 Special Files for Network Communications

Starting with version 3.1 of gawk, awk programs can open a two-way TCP/IP connection, acting as either a client or a server. This is done using a special file name of the form:


The protocol is one of ‘tcp’, ‘udp’, or ‘raw’, and the other fields represent the other essential pieces of information for making a networking connection. These file names are used with the ‘|&’ operator for communicating with a coprocess (see Two-way I/O). This is an advanced feature, mentioned here only for completeness. Full discussion is delayed until TCP/IP Networking.

Previous: Special Network, Up: Special Files

4.7.4 Special File Name Caveats

Here is a list of things to bear in mind when using the special file names that gawk provides:

Previous: Special Files, Up: Printing

4.8 Closing Input and Output Redirections

If the same file name or the same shell command is used with getline more than once during the execution of an awk program (see Getline), the file is opened (or the command is executed) the first time only. At that time, the first record of input is read from that file or command. The next time the same file or command is used with getline, another record is read from it, and so on.

Similarly, when a file or pipe is opened for output, the file name or command associated with it is remembered by awk, and subsequent writes to the same file or command are appended to the previous writes. The file or pipe stays open until awk exits.

This implies that special steps are necessary in order to read the same file again from the beginning, or to rerun a shell command (rather than reading more output from the same command). The close function makes these things possible:




The argument filename or command can be any expression. Its value must exactly match the string that was used to open the file or start the command (spaces and other “irrelevant” characters included). For example, if you open a pipe with this:

     "sort -r names" | getline foo

then you must close it with this:

     close("sort -r names")

Once this function call is executed, the next getline from that file or command, or the next print or printf to that file or command, reopens the file or reruns the command. Because the expression that you use to close a file or pipeline must exactly match the expression used to open the file or run the command, it is good practice to use a variable to store the file name or command. The previous example becomes the following:

     sortcom = "sort -r names"
     sortcom | getline foo

This helps avoid hard-to-find typographical errors in your awk programs. Here are some of the reasons for closing an output file:

If you use more files than the system allows you to have open, gawk attempts to multiplex the available open files among your data files. gawk's ability to do this depends upon the facilities of your operating system, so it may not always work. It is therefore both good practice and good portability advice to always use close on your files when you are done with them. In fact, if you are using a lot of pipes, it is essential that you close commands when done. For example, consider something like this:

         command = ("grep " $1 " /some/file | my_prog -q " $3)
         while ((command | getline) > 0) {
             process output of command
         # need close(command) here

This example creates a new pipeline based on data in each record. Without the call to close indicated in the comment, awk creates child processes to run the commands, until it eventually runs out of file descriptors for more pipelines.

Even though each command has finished (as indicated by the end-of-file return status from getline), the child process is not terminated;21 more importantly, the file descriptor for the pipe is not closed and released until close is called or awk exits.

close will silently do nothing if given an argument that does not represent a file, pipe or coprocess that was opened with a redirection.

Note also that ‘close(FILENAME)’ has no “magic” effects on the implicit loop that reads through the files named on the command line. It is, more likely, a close of a file that was never opened, so awk silently does nothing.

When using the ‘|&’ operator to communicate with a coprocess, it is occasionally useful to be able to close one end of the two-way pipe without closing the other. This is done by supplying a second argument to close. As in any other call to close, the first argument is the name of the command or special file used to start the coprocess. The second argument should be a string, with either of the values "to" or "from". Case does not matter. As this is an advanced feature, a more complete discussion is delayed until Two-way I/O, which discusses it in more detail and gives an example.

Advanced Notes: Using close's Return Value

In many versions of Unix awk, the close function is actually a statement. It is a syntax error to try and use the return value from close: (d.c.)

     command = "..."
     command | getline info
     retval = close(command)  # syntax error in most Unix awks

gawk treats close as a function. The return value is −1 if the argument names something that was never opened with a redirection, or if there is a system problem closing the file or process. In these cases, gawk sets the built-in variable ERRNO to a string describing the problem.

In gawk, when closing a pipe or coprocess (input or output), the return value is the exit status of the command.22 Otherwise, it is the return value from the system's close or fclose C functions when closing input or output files, respectively. This value is zero if the close succeeds, or −1 if it fails.

The POSIX standard is very vague; it says that close returns zero on success and non-zero otherwise. In general, different implementations vary in what they report when closing pipes; thus the return value cannot be used portably. (d.c.)

Next: , Previous: Printing, Up: Top

5 Expressions

Expressions are the basic building blocks of awk patterns and actions. An expression evaluates to a value that you can print, test, or pass to a function. Additionally, an expression can assign a new value to a variable or a field by using an assignment operator.

An expression can serve as a pattern or action statement on its own. Most other kinds of statements contain one or more expressions that specify the data on which to operate. As in other languages, expressions in awk include variables, array references, constants, and function calls, as well as combinations of these with various operators.

Next: , Up: Expressions

5.1 Constant Expressions

The simplest type of expression is the constant, which always has the same value. There are three types of constants: numeric, string, and regular expression.

Each is used in the appropriate context when you need a data value that isn't going to change. Numeric constants can have different forms, but are stored identically internally.

Next: , Up: Constants

5.1.1 Numeric and String Constants

A numeric constant stands for a number. This number can be an integer, a decimal fraction, or a number in scientific (exponential) notation.23 Here are some examples of numeric constants that all have the same value:


A string constant consists of a sequence of characters enclosed in double-quotation marks. For example:


represents the string whose contents are ‘parrot’. Strings in gawk can be of any length, and they can contain any of the possible eight-bit ASCII characters including ASCII nul (character code zero). Other awk implementations may have difficulty with some character codes.

Next: , Previous: Scalar Constants, Up: Constants

5.1.2 Octal and Hexadecimal Numbers

In awk, all numbers are in decimal; i.e., base 10. Many other programming languages allow you to specify numbers in other bases, often octal (base 8) and hexadecimal (base 16). In octal, the numbers go 0, 1, 2, 3, 4, 5, 6, 7, 10, 11, 12, etc. Just as ‘11’, in decimal, is 1 times 10 plus 1, so ‘11’, in octal, is 1 times 8, plus 1. This equals 9 in decimal. In hexadecimal, there are 16 digits. Since the everyday decimal number system only has ten digits (‘0’–‘9’), the letters ‘a’ through ‘f’ are used to represent the rest. (Case in the letters is usually irrelevant; hexadecimal ‘a’ and ‘A’ have the same value.) Thus, ‘11’, in hexadecimal, is 1 times 16 plus 1, which equals 17 in decimal.

Just by looking at plain ‘11’, you can't tell what base it's in. So, in C, C++, and other languages derived from C, there is a special notation to help signify the base. Octal numbers start with a leading ‘0’, and hexadecimal numbers start with a leading ‘0x’ or ‘0X’:

Decimal value 11.
Octal 11, decimal value 9.
Hexadecimal 11, decimal value 17.

This example shows the difference:

     $ gawk 'BEGIN { printf "%d, %d, %d\n", 011, 11, 0x11 }'
     -| 9, 11, 17

Being able to use octal and hexadecimal constants in your programs is most useful when working with data that cannot be represented conveniently as characters or as regular numbers, such as binary data of various sorts.

gawk allows the use of octal and hexadecimal constants in your program text. However, such numbers in the input data are not treated differently; doing so by default would break old programs. (If you really need to do this, use the --non-decimal-data command-line option; see Nondecimal Data.) If you have octal or hexadecimal data, you can use the strtonum function (see String Functions) to convert the data into a number. Most of the time, you will want to use octal or hexadecimal constants when working with the built-in bit manipulation functions; see Bitwise Functions, for more information.

Unlike some early C implementations, ‘8’ and ‘9’ are not valid in octal constants; e.g., gawk treats ‘018’ as decimal 18:

     $ gawk 'BEGIN { print "021 is", 021 ; print 018 }'
     -| 021 is 17
     -| 18

Octal and hexadecimal source code constants are a gawk extension. If gawk is in compatibility mode (see Options), they are not available.

Advanced Notes: A Constant's Base Does Not Affect Its Value

Once a numeric constant has been converted internally into a number, gawk no longer remembers what the original form of the constant was; the internal value is always used. This has particular consequences for conversion of numbers to strings:

     $ gawk 'BEGIN { printf "0x11 is <%s>\n", 0x11 }'
     -| 0x11 is <17>

Previous: Nondecimal-numbers, Up: Constants

5.1.3 Regular Expression Constants

A regexp constant is a regular expression description enclosed in slashes, such as /^beginning and end$/. Most regexps used in awk programs are constant, but the ‘~’ and ‘!~’ matching operators can also match computed or “dynamic” regexps (which are just ordinary strings or variables that contain a regexp).

Next: , Previous: Constants, Up: Expressions

5.2 Using Regular Expression Constants

When used on the righthand side of the ‘~’ or ‘!~’ operators, a regexp constant merely stands for the regexp that is to be matched. However, regexp constants (such as /foo/) may be used like simple expressions. When a regexp constant appears by itself, it has the same meaning as if it appeared in a pattern, i.e., ‘($0 ~ /foo/)’ (d.c.) See Expression Patterns. This means that the following two code segments:

     if ($0 ~ /barfly/ || $0 ~ /camelot/)
         print "found"


     if (/barfly/ || /camelot/)
         print "found"

are exactly equivalent. One rather bizarre consequence of this rule is that the following Boolean expression is valid, but does not do what the user probably intended:

     # note that /foo/ is on the left of the ~
     if (/foo/ ~ $1) print "found foo"

This code is “obviously” testing $1 for a match against the regexp /foo/. But in fact, the expression ‘/foo/ ~ $1’ actually means ‘($0 ~ /foo/) ~ $1’. In other words, first match the input record against the regexp /foo/. The result is either zero or one, depending upon the success or failure of the match. That result is then matched against the first field in the record. Because it is unlikely that you would ever really want to make this kind of test, gawk issues a warning when it sees this construct in a program. Another consequence of this rule is that the assignment statement:

     matches = /foo/

assigns either zero or one to the variable matches, depending upon the contents of the current input record. This feature of the language has never been well documented until the POSIX specification.

Constant regular expressions are also used as the first argument for the gensub, sub, and gsub functions, and as the second argument of the match function (see String Functions). Modern implementations of awk, including gawk, allow the third argument of split to be a regexp constant, but some older implementations do not. (d.c.) This can lead to confusion when attempting to use regexp constants as arguments to user-defined functions (see User-defined). For example:

     function mysub(pat, repl, str, global)
         if (global)
             gsub(pat, repl, str)
             sub(pat, repl, str)
         return str
         text = "hi! hi yourself!"
         mysub(/hi/, "howdy", text, 1)

In this example, the programmer wants to pass a regexp constant to the user-defined function mysub, which in turn passes it on to either sub or gsub. However, what really happens is that the pat parameter is either one or zero, depending upon whether or not $0 matches /hi/. gawk issues a warning when it sees a regexp constant used as a parameter to a user-defined function, since passing a truth value in this way is probably not what was intended.

Next: , Previous: Using Constant Regexps, Up: Expressions

5.3 Variables

Variables are ways of storing values at one point in your program for use later in another part of your program. They can be manipulated entirely within the program text, and they can also be assigned values on the awk command line.

Next: , Up: Variables

5.3.1 Using Variables in a Program

Variables let you give names to values and refer to them later. Variables have already been used in many of the examples. The name of a variable must be a sequence of letters, digits, or underscores, and it may not begin with a digit. Case is significant in variable names; a and A are distinct variables.

A variable name is a valid expression by itself; it represents the variable's current value. Variables are given new values with assignment operators, increment operators, and decrement operators. See Assignment Ops.

A few variables have special built-in meanings, such as FS (the field separator), and NF (the number of fields in the current input record). See Built-in Variables, for a list of the built-in variables. These built-in variables can be used and assigned just like all other variables, but their values are also used or changed automatically by awk. All built-in variables' names are entirely uppercase.

Variables in awk can be assigned either numeric or string values. The kind of value a variable holds can change over the life of a program. By default, variables are initialized to the empty string, which is zero if converted to a number. There is no need to “initialize” each variable explicitly in awk, which is what you would do in C and in most other traditional languages.

Previous: Using Variables, Up: Variables

5.3.2 Assigning Variables on the Command Line

Any awk variable can be set by including a variable assignment among the arguments on the command line when awk is invoked (see Other Arguments). Such an assignment has the following form:


With it, a variable is set either at the beginning of the awk run or in between input files. When the assignment is preceded with the -v option, as in the following:

     -v variable=text

the variable is set at the very beginning, even before the BEGIN rules are run. The -v option and its assignment must precede all the file name arguments, as well as the program text. (See Options, for more information about the -v option.) Otherwise, the variable assignment is performed at a time determined by its position among the input file arguments—after the processing of the preceding input file argument. For example:

     awk '{ print $n }' n=4 inventory-shipped n=2 BBS-list

prints the value of field number n for all input records. Before the first file is read, the command line sets the variable n equal to four. This causes the fourth field to be printed in lines from the file inventory-shipped. After the first file has finished, but before the second file is started, n is set to two, so that the second field is printed in lines from BBS-list:

     $ awk '{ print $n }' n=4 inventory-shipped n=2 BBS-list
     -| 15
     -| 24
     -| 555-5553
     -| 555-3412

Command-line arguments are made available for explicit examination by the awk program in the ARGV array (see ARGC and ARGV). awk processes the values of command-line assignments for escape sequences (see Escape Sequences). (d.c.)

Next: , Previous: Variables, Up: Expressions

5.4 Conversion of Strings and Numbers

Strings are converted to numbers and numbers are converted to strings, if the context of the awk program demands it. For example, if the value of either foo or bar in the expression ‘foo + bar’ happens to be a string, it is converted to a number before the addition is performed. If numeric values appear in string concatenation, they are converted to strings. Consider the following:

     two = 2; three = 3
     print (two three) + 4

This prints the (numeric) value 27. The numeric values of the variables two and three are converted to strings and concatenated together. The resulting string is converted back to the number 23, to which 4 is then added.

If, for some reason, you need to force a number to be converted to a string, concatenate the empty string, "", with that number. To force a string to be converted to a number, add zero to that string. A string is converted to a number by interpreting any numeric prefix of the string as numerals: "2.5" converts to 2.5, "1e3" converts to 1000, and "25fix" has a numeric value of 25. Strings that can't be interpreted as valid numbers convert to zero.

The exact manner in which numbers are converted into strings is controlled by the awk built-in variable CONVFMT (see Built-in Variables). Numbers are converted using the sprintf function with CONVFMT as the format specifier (see String Functions).

CONVFMT's default value is "%.6g", which prints a value with at least six significant digits. For some applications, you might want to change it to specify more precision. On most modern machines, 17 digits is enough to capture a floating-point number's value exactly, most of the time.24

Strange results can occur if you set CONVFMT to a string that doesn't tell sprintf how to format floating-point numbers in a useful way. For example, if you forget the ‘%’ in the format, awk converts all numbers to the same constant string. As a special case, if a number is an integer, then the result of converting it to a string is always an integer, no matter what the value of CONVFMT may be. Given the following code fragment:

     CONVFMT = "%2.2f"
     a = 12
     b = a ""

b has the value "12", not "12.00". (d.c.)

Prior to the POSIX standard, awk used the value of OFMT for converting numbers to strings. OFMT specifies the output format to use when printing numbers with print. CONVFMT was introduced in order to separate the semantics of conversion from the semantics of printing. Both CONVFMT and OFMT have the same default value: "%.6g". In the vast majority of cases, old awk programs do not change their behavior. However, these semantics for OFMT are something to keep in mind if you must port your new style program to older implementations of awk. We recommend that instead of changing your programs, just port gawk itself. See Print, for more information on the print statement.

And, once again, where you are can matter when it comes to converting between numbers and strings. In Locales, we mentioned that the local character set and language (the locale) can affect how gawk matches characters. The locale also affects numeric formats. In particular, for awk programs, it affects the decimal point character. The "C" locale, and most English-language locales, use the period character (‘.’) as the decimal point. However, many (if not most) European and non-English locales use the comma (‘,’) as the decimal point character.

The POSIX standard says that awk always uses the period as the decimal point when reading the awk program source code, and for command-line variable assignments (see Other Arguments). However, when interpreting input data, for print and printf output, and for number to string conversion, the local decimal point character is used. Here are some examples indicating the difference in behavior, on a GNU/Linux system:

     $ gawk 'BEGIN { printf "%g\n", 3.1415927 }'
     -| 3.14159
     $  LC_ALL=en_DK gawk 'BEGIN { printf "%g\n", 3.1415927 }'
     -| 3,14159
     $ echo 4,321 | gawk '{ print $1 + 1 }'
     -| 5
     $ echo 4,321 | LC_ALL=en_DK gawk '{ print $1 + 1 }'
     -| 5,321

The ‘en_DK’ locale is for English in Denmark, where the comma acts as the decimal point separator. In the normal "C" locale, gawk treats ‘4,321’ as ‘4’, while in the Danish locale, it's treated as the full number, ‘4.321’.

For version 3.1.3 through 3.1.5, gawk fully complied with this aspect of the standard. However, many users in non-English locales complained about this behavior, since their data used a period as the decimal point. Beginning in version 3.1.6, the default behavior was restored to use a period as the decimal point character. You can use the --use-lc-numeric option (see Options) to force gawk to use the locale's decimal point character. (gawk also uses the locale's decimal point character when in POSIX mode, either via --posix, or the POSIXLY_CORRECT environment variable.)

The following table describes the cases in which the locale's decimal point character is used and when a period is used. Some of these features have not been described yet.

Feature Default --posix or --use-lc-numeric
%'gUse locale Use locale
%gUse period Use locale
Input Use period Use locale
strtonumUse period Use locale

Table 5.1: Locale Decimal Point versus A Period

Finally, modern day formal standards and IEEE standard floating point representation can have an unusual but important effect on the way gawk converts some special string values to numbers. The details are presented in POSIX Floating Point Problems.

Next: , Previous: Conversion, Up: Expressions

5.5 Arithmetic Operators

The awk language uses the common arithmetic operators when evaluating expressions. All of these arithmetic operators follow normal precedence rules and work as you would expect them to.

The following example uses a file named grades, which contains a list of student names as well as three test scores per student (it's a small class):

     Pat   100 97 58
     Sandy  84 72 93
     Chris  72 92 89

This programs takes the file grades and prints the average of the scores:

     $ awk '{ sum = $2 + $3 + $4 ; avg = sum / 3
     >        print $1, avg }' grades
     -| Pat 85
     -| Sandy 83
     -| Chris 84.3333

The following list provides the arithmetic operators in awk, in order from the highest precedence to the lowest:

- x
+ x
Unary plus; the expression is converted to a number.

x ^ y
x ** y
Exponentiation; x raised to the y power. ‘2 ^ 3’ has the value eight; the character sequence ‘**’ is equivalent to ‘^’.
x * y

x / y
Division; because all numbers in awk are floating-point numbers, the result is not rounded to an integer—‘3 / 4’ has the value 0.75. (It is a common mistake, especially for C programmers, to forget that all numbers in awk are floating-point, and that division of integer-looking constants produces a real number, not an integer.)
x % y
Remainder; further discussion is provided in the text, just after this list.
x + y
x - y

Unary plus and minus have the same precedence, the multiplication operators all have the same precedence, and addition and subtraction have the same precedence.

When computing the remainder of x % y, the quotient is rounded toward zero to an integer and multiplied by y. This result is subtracted from x; this operation is sometimes known as “trunc-mod.” The following relation always holds:

     b * int(a / b) + (a % b) == a

One possibly undesirable effect of this definition of remainder is that x % y is negative if x is negative. Thus:

     -17 % 8 = -1

In other awk implementations, the signedness of the remainder may be machine-dependent.

NOTE: The POSIX standard only specifies the use of ‘^’ for exponentiation. For maximum portability, do not use the ‘**’ operator.

Next: , Previous: Arithmetic Ops, Up: Expressions

5.6 String Concatenation

It seemed like a good idea at the time.
Brian Kernighan

There is only one string operation: concatenation. It does not have a specific operator to represent it. Instead, concatenation is performed by writing expressions next to one another, with no operator. For example:

     $ awk '{ print "Field number one: " $1 }' BBS-list
     -| Field number one: aardvark
     -| Field number one: alpo-net

Without the space in the string constant after the ‘:’, the line runs together. For example:

     $ awk '{ print "Field number one:" $1 }' BBS-list
     -| Field number one:aardvark
     -| Field number one:alpo-net

Because string concatenation does not have an explicit operator, it is often necessary to insure that it happens at the right time by using parentheses to enclose the items to concatenate. For example, you might expect that the following code fragment concatenates file and name:

     file = "file"
     name = "name"
     print "something meaningful" > file name

This produces a syntax error with Unix awk.25 It is necessary to use the following:

     print "something meaningful" > (file name)

Parentheses should be used around concatenation in all but the most common contexts, such as on the righthand side of ‘=’. Be careful about the kinds of expressions used in string concatenation. In particular, the order of evaluation of expressions used for concatenation is undefined in the awk language. Consider this example:

     BEGIN {
         a = "don't"
         print (a " " (a = "panic"))

It is not defined whether the assignment to a happens before or after the value of a is retrieved for producing the concatenated value. The result could be either ‘don't panic’, or ‘panic panic’. The precedence of concatenation, when mixed with other operators, is often counter-intuitive. Consider this example:

     $ awk 'BEGIN { print -12 " " -24 }'
     -| -12-24

This “obviously” is concatenating −12, a space, and −24. But where did the space disappear to? The answer lies in the combination of operator precedences and awk's automatic conversion rules. To get the desired result, write the program in the following manner:

     $ awk 'BEGIN { print -12 " " (-24) }'
     -| -12 -24

This forces awk to treat the ‘-’ on the ‘-24’ as unary. Otherwise, it's parsed as follows:

         −12 (" " − 24)
     ⇒ −12 (0 − 24)
     ⇒ −12 (−24)
     ⇒ −12−24

As mentioned earlier, when doing concatenation, parenthesize. Otherwise, you're never quite sure what you'll get.

Next: , Previous: Concatenation, Up: Expressions

5.7 Assignment Expressions

An assignment is an expression that stores a (usually different) value into a variable. For example, let's assign the value one to the variable z:

     z = 1

After this expression is executed, the variable z has the value one. Whatever old value z had before the assignment is forgotten.

Assignments can also store string values. For example, the following stores the value "this food is good" in the variable message:

     thing = "food"
     predicate = "good"
     message = "this " thing " is " predicate

This also illustrates string concatenation. The ‘=’ sign is called an assignment operator. It is the simplest assignment operator because the value of the righthand operand is stored unchanged. Most operators (addition, concatenation, and so on) have no effect except to compute a value. If the value isn't used, there's no reason to use the operator. An assignment operator is different; it does produce a value, but even if you ignore it, the assignment still makes itself felt through the alteration of the variable. We call this a side effect.

The lefthand operand of an assignment need not be a variable (see Variables); it can also be a field (see Changing Fields) or an array element (see Arrays). These are all called lvalues, which means they can appear on the lefthand side of an assignment operator. The righthand operand may be any expression; it produces the new value that the assignment stores in the specified variable, field, or array element. (Such values are called rvalues.)

It is important to note that variables do not have permanent types. A variable's type is simply the type of whatever value it happens to hold at the moment. In the following program fragment, the variable foo has a numeric value at first, and a string value later on:

     foo = 1
     print foo
     foo = "bar"
     print foo

When the second assignment gives foo a string value, the fact that it previously had a numeric value is forgotten.

String values that do not begin with a digit have a numeric value of zero. After executing the following code, the value of foo is five:

     foo = "a string"
     foo = foo + 5
NOTE: Using a variable as a number and then later as a string can be confusing and is poor programming style. The previous two examples illustrate how awk works, not how you should write your programs!

An assignment is an expression, so it has a value—the same value that is assigned. Thus, ‘z = 1’ is an expression with the value one. One consequence of this is that you can write multiple assignments together, such as:

     x = y = z = 5

This example stores the value five in all three variables (x, y, and z). It does so because the value of ‘z = 5’, which is five, is stored into y and then the value of ‘y = z = 5’, which is five, is stored into x.

Assignments may be used anywhere an expression is called for. For example, it is valid to write ‘x != (y = 1)’ to set y to one, and then test whether x equals one. But this style tends to make programs hard to read; such nesting of assignments should be avoided, except perhaps in a one-shot program.

Aside from ‘=’, there are several other assignment operators that do arithmetic with the old value of the variable. For example, the operator ‘+=’ computes a new value by adding the righthand value to the old value of the variable. Thus, the following assignment adds five to the value of foo:

     foo += 5

This is equivalent to the following:

     foo = foo + 5

Use whichever makes the meaning of your program clearer.

There are situations where using ‘+=’ (or any assignment operator) is not the same as simply repeating the lefthand operand in the righthand expression. For example:

     # Thanks to Pat Rankin for this example
     BEGIN  {
         foo[rand()] += 5
         for (x in foo)
            print x, foo[x]
         bar[rand()] = bar[rand()] + 5
         for (x in bar)
            print x, bar[x]

The indices of bar are practically guaranteed to be different, because rand returns different values each time it is called. (Arrays and the rand function haven't been covered yet. See Arrays, and see Numeric Functions, for more information). This example illustrates an important fact about assignment operators: the lefthand expression is only evaluated once. It is up to the implementation as to which expression is evaluated first, the lefthand or the righthand. Consider this example:

     i = 1
     a[i += 2] = i + 1

The value of a[3] could be either two or four.

table-assign-ops lists the arithmetic assignment operators. In each case, the righthand operand is an expression whose value is converted to a number.

Operator Effect
lvalue += increment Adds increment to the value of lvalue.
lvalue -= decrement Subtracts decrement from the value of lvalue.
lvalue *= coefficient Multiplies the value of lvalue by coefficient.
lvalue /= divisor Divides the value of lvalue by divisor.
lvalue %= modulus Sets lvalue to its remainder by modulus.
lvalue ^= power
lvalue **= power Raises lvalue to the power power.

Table 5.2: Arithmetic Assignment Operators

NOTE: Only the ‘^=’ operator is specified by POSIX. For maximum portability, do not use the ‘**=’ operator.

Advanced Notes: Syntactic Ambiguities Between ‘/=’ and Regular Expressions

There is a syntactic ambiguity between the ‘/=’ assignment operator and regexp constants whose first character is an ‘=’. (d.c.) This is most notable in commercial awk versions. For example:

     $ awk /==/ /dev/null
     error--> awk: syntax error at source line 1
     error-->  context is
     error-->         >>> /= <<<
     error--> awk: bailing out at source line 1

A workaround is:

     awk '/[=]=/' /dev/null

gawk does not have this problem, nor do the other freely available versions described in Other Versions.

Next: , Previous: Assignment Ops, Up: Expressions

5.8 Increment and Decrement Operators

Increment and decrement operators increase or decrease the value of a variable by one. An assignment operator can do the same thing, so the increment operators add no power to the awk language; however, they are convenient abbreviations for very common operations.

The operator used for adding one is written ‘++’. It can be used to increment a variable either before or after taking its value. To pre-increment a variable v, write ‘++v’. This adds one to the value of v—that new value is also the value of the expression. (The assignment expression ‘v += 1’ is completely equivalent.) Writing the ‘++’ after the variable specifies post-increment. This increments the variable value just the same; the difference is that the value of the increment expression itself is the variable's old value. Thus, if foo has the value four, then the expression ‘foo++’ has the value four, but it changes the value of foo to five. In other words, the operator returns the old value of the variable, but with the side effect of incrementing it.

The post-increment ‘foo++’ is nearly the same as writing ‘(foo += 1) - 1’. It is not perfectly equivalent because all numbers in awk are floating-point—in floating-point, ‘foo + 1 - 1’ does not necessarily equal foo. But the difference is minute as long as you stick to numbers that are fairly small (less than 10e12).

Fields and array elements are incremented just like variables. (Use ‘$(i++)’ when you want to do a field reference and a variable increment at the same time. The parentheses are necessary because of the precedence of the field reference operator ‘$’.)

The decrement operator ‘--’ works just like ‘++’, except that it subtracts one instead of adding it. As with ‘++’, it can be used before the lvalue to pre-decrement or after it to post-decrement. Following is a summary of increment and decrement expressions:

This expression increments lvalue, and the new value becomes the value of the expression.
This expression increments lvalue, but the value of the expression is the old value of lvalue.

This expression is like ‘++lvalue’, but instead of adding, it subtracts. It decrements lvalue and delivers the value that is the result.
This expression is like ‘lvalue++’, but instead of adding, it subtracts. It decrements lvalue. The value of the expression is the old value of lvalue.

Advanced Notes: Operator Evaluation Order

Doctor, doctor! It hurts when I do this!
So don't do that!

Groucho Marx

What happens for something like the following?

     b = 6
     print b += b++

Or something even stranger?

     b = 6
     b += ++b + b++
     print b

In other words, when do the various side effects prescribed by the postfix operators (‘b++’) take effect? When side effects happen is implementation defined. In other words, it is up to the particular version of awk. The result for the first example may be 12 or 13, and for the second, it may be 22 or 23.

In short, doing things like this is not recommended and definitely not anything that you can rely upon for portability. You should avoid such things in your own programs.

Next: , Previous: Increment Ops, Up: Expressions

5.9 True and False in awk

Many programming languages have a special representation for the concepts of “true” and “false.” Such languages usually use the special constants true and false, or perhaps their uppercase equivalents. However, awk is different. It borrows a very simple concept of true and false from C. In awk, any nonzero numeric value or any nonempty string value is true. Any other value (zero or the null string "") is false. The following program prints ‘A strange truth value’ three times:

     BEGIN {
        if (3.1415927)
            print "A strange truth value"
        if ("Four Score And Seven Years Ago")
            print "A strange truth value"
        if (j = 57)
            print "A strange truth value"

There is a surprising consequence of the “nonzero or non-null” rule: the string constant "0" is actually true, because it is non-null. (d.c.)

Next: , Previous: Truth Values, Up: Expressions

5.10 Variable Typing and Comparison Expressions

The Guide is definitive. Reality is frequently inaccurate.
The Hitchhiker's Guide to the Galaxy

Unlike other programming languages, awk variables do not have a fixed type. Instead, they can be either a number or a string, depending upon the value that is assigned to them. We look now at how variables are typed, and how awk compares variables.

Next: , Up: Typing and Comparison

5.10.1 String Type Versus Numeric Type

The 1992 POSIX standard introduced the concept of a numeric string, which is simply a string that looks like a number—for example, " +2". This concept is used for determining the type of a variable. The type of the variable is important because the types of two variables determine how they are compared. In gawk, variable typing follows these rules:

The last rule is particularly important. In the following program, a has numeric type, even though it is later used in a string operation:

     BEGIN {
          a = 12.345
          b = a " is a cute number"
          print b

When two operands are compared, either string comparison or numeric comparison may be used. This depends upon the attributes of the operands, according to the following symmetric matrix:

             |       STRING          NUMERIC         STRNUM
     STRING  |       string          string          string
     NUMERIC |       string          numeric         numeric
     STRNUM  |       string          numeric         numeric

The basic idea is that user input that looks numeric—and only user input—should be treated as numeric, even though it is actually made of characters and is therefore also a string. Thus, for example, the string constant " +3.14", when it appears in program source code, is a string—even though it looks numeric—and is never treated as number for comparison purposes.

In short, when one operand is a “pure” string, such as a string constant, then a string comparison is performed. Otherwise, a numeric comparison is performed.26

This point bears additional emphasis: All user input is made of characters, and so is first and foremost of string type; input strings that look numeric are additionally given the strnum attribute. Thus, the six-character input string ‘ +3.14 receives the strnum attribute. In contrast, the eight-character literal " +3.14" appearing in program text is a string constant. The following examples print ‘1’ when the comparison between the two different constants is true, ‘0’ otherwise:

     $ echo ' +3.14' | gawk '{ print $0 == " +3.14" }'    True
     -| 1
     $ echo ' +3.14' | gawk '{ print $0 == "+3.14" }'     False
     -| 0
     $ echo ' +3.14' | gawk '{ print $0 == "3.14" }'      False
     -| 0
     $ echo ' +3.14' | gawk '{ print $0 == 3.14 }'        True
     -| 1
     $ echo ' +3.14' | gawk '{ print $1 == " +3.14" }'    False
     -| 0
     $ echo ' +3.14' | gawk '{ print $1 == "+3.14" }'     True
     -| 1
     $ echo ' +3.14' | gawk '{ print $1 == "3.14" }'      False
     -| 0
     $ echo ' +3.14' | gawk '{ print $1 == 3.14 }'        True
     -| 1

Previous: Variable Typing, Up: Typing and Comparison

5.10.2 Comparison Operators

Comparison expressions compare strings or numbers for relationships such as equality. They are written using relational operators, which are a superset of those in C. table-relational-ops describes them.

Expression Result
x < y True if x is less than y.
x <= y True if x is less than or equal to y.
x > y True if x is greater than y.
x >= y True if x is greater than or equal to y.
x == y True if x is equal to y.
x != y True if x is not equal to y.
x ~ y True if the string x matches the regexp denoted by y.
x !~ y True if the string x does not match the regexp denoted by y.
subscript in array True if the array array has an element with the subscript subscript.

Table 5.3: Relational Operators

Comparison expressions have the value one if true and zero if false. When comparing operands of mixed types, numeric operands are converted to strings using the value of CONVFMT (see Conversion).

Strings are compared by comparing the first character of each, then the second character of each, and so on. Thus, "10" is less than "9". If there are two strings where one is a prefix of the other, the shorter string is less than the longer one. Thus, "abc" is less than "abcd".

It is very easy to accidentally mistype the ‘==’ operator and leave off one of the ‘=’ characters. The result is still valid awk code, but the program does not do what is intended:

     if (a = b)   # oops! should be a == b

Unless b happens to be zero or the null string, the if part of the test always succeeds. Because the operators are so similar, this kind of error is very difficult to spot when scanning the source code.

The following table of expressions illustrates the kind of comparison gawk performs, as well as what the result of the comparison is:

1.5 <= 2.0
numeric comparison (true)
"abc" >= "xyz"
string comparison (false)
1.5 != " +2"
string comparison (true)
"1e2" < "3"
string comparison (true)
a = 2; b = "2"
a == b
string comparison (true)
a = 2; b = " +2"
a == b
string comparison (false)

In the next example:

     $ echo 1e2 3 | awk '{ print ($1 < $2) ? "true" : "false" }'
     -| false

the result is ‘false’ because both $1 and $2 are user input. They are numeric strings—therefore both have the strnum attribute, dictating a numeric comparison. The purpose of the comparison rules and the use of numeric strings is to attempt to produce the behavior that is “least surprising,” while still “doing the right thing.” String comparisons and regular expression comparisons are very different. For example:

     x == "foo"

has the value one, or is true if the variable x is precisely ‘foo’. By contrast:

     x ~ /foo/

has the value one if x contains ‘foo’, such as "Oh, what a fool am I!".

The righthand operand of the ‘~’ and ‘!~’ operators may be either a regexp constant (/.../) or an ordinary expression. In the latter case, the value of the expression as a string is used as a dynamic regexp (see Regexp Usage; also see Computed Regexps).

In modern implementations of awk, a constant regular expression in slashes by itself is also an expression. The regexp /regexp/ is an abbreviation for the following comparison expression:

     $0 ~ /regexp/

One special place where /foo/ is not an abbreviation for ‘$0 ~ /foo/’ is when it is the righthand operand of ‘~’ or ‘!~’. See Using Constant Regexps, where this is discussed in more detail.

Next: , Previous: Typing and Comparison, Up: Expressions

5.11 Boolean Expressions

A Boolean expression is a combination of comparison expressions or matching expressions, using the Boolean operators “or” (‘||’), “and” (‘&&’), and “not” (‘!’), along with parentheses to control nesting. The truth value of the Boolean expression is computed by combining the truth values of the component expressions. Boolean expressions are also referred to as logical expressions. The terms are equivalent.

Boolean expressions can be used wherever comparison and matching expressions can be used. They can be used in if, while, do, and for statements (see Statements). They have numeric values (one if true, zero if false) that come into play if the result of the Boolean expression is stored in a variable or used in arithmetic.

In addition, every Boolean expression is also a valid pattern, so you can use one as a pattern to control the execution of rules. The Boolean operators are:

boolean1 && boolean2
True if both boolean1 and boolean2 are true. For example, the following statement prints the current input record if it contains both ‘2400’ and ‘foo’:
          if ($0 ~ /2400/ && $0 ~ /foo/) print

The subexpression boolean2 is evaluated only if boolean1 is true. This can make a difference when boolean2 contains expressions that have side effects. In the case of ‘$0 ~ /foo/ && ($2 == bar++)’, the variable bar is not incremented if there is no substring ‘foo’ in the record.

boolean1 || boolean2
True if at least one of boolean1 or boolean2 is true. For example, the following statement prints all records in the input that contain either2400’ or ‘foo’ or both:
          if ($0 ~ /2400/ || $0 ~ /foo/) print

The subexpression boolean2 is evaluated only if boolean1 is false. This can make a difference when boolean2 contains expressions that have side effects.

! boolean
True if boolean is false. For example, the following program prints ‘no home!’ in the unusual event that the HOME environment variable is not defined:
          BEGIN { if (! ("HOME" in ENVIRON))
                         print "no home!" }

(The in operator is described in Reference to Elements.)

The ‘&&’ and ‘||’ operators are called short-circuit operators because of the way they work. Evaluation of the full expression is “short-circuited” if the result can be determined part way through its evaluation.

Statements that use ‘&&’ or ‘||’ can be continued simply by putting a newline after them. But you cannot put a newline in front of either of these operators without using backslash continuation (see Statements/Lines).

The actual value of an expression using the ‘!’ operator is either one or zero, depending upon the truth value of the expression it is applied to. The ‘!’ operator is often useful for changing the sense of a flag variable from false to true and back again. For example, the following program is one way to print lines in between special bracketing lines:

     $1 == "START"   { interested = ! interested; next }
     interested == 1 { print }
     $1 == "END"     { interested = ! interested; next }

The variable interested, as with all awk variables, starts out initialized to zero, which is also false. When a line is seen whose first field is ‘START’, the value of interested is toggled to true, using ‘!’. The next rule prints lines as long as interested is true. When a line is seen whose first field is ‘END’, interested is toggled back to false.27

NOTE: The next statement is discussed in Next Statement. next tells awk to skip the rest of the rules, get the next record, and start processing the rules over again at the top. The reason it's there is to avoid printing the bracketing ‘START’ and ‘END’ lines.

Next: , Previous: Boolean Ops, Up: Expressions

5.12 Conditional Expressions

A conditional expression is a special kind of expression that has three operands. It allows you to use one expression's value to select one of two other expressions. The conditional expression is the same as in the C language, as shown here:

     selector ? if-true-exp : if-false-exp

There are three subexpressions. The first, selector, is always computed first. If it is “true” (not zero or not null), then if-true-exp is computed next and its value becomes the value of the whole expression. Otherwise, if-false-exp is computed next and its value becomes the value of the whole expression. For example, the following expression produces the absolute value of x:

     x >= 0 ? x : -x

Each time the conditional expression is computed, only one of if-true-exp and if-false-exp is used; the other is ignored. This is important when the expressions have side effects. For example, this conditional expression examines element i of either array a or array b, and increments i:

     x == y ? a[i++] : b[i++]

This is guaranteed to increment i exactly once, because each time only one of the two increment expressions is executed and the other is not. See Arrays, for more information about arrays.

As a minor gawk extension, a statement that uses ‘?:’ can be continued simply by putting a newline after either character. However, putting a newline in front of either character does not work without using backslash continuation (see Statements/Lines). If --posix is specified (see Options), then this extension is disabled.

Next: , Previous: Conditional Exp, Up: Expressions

5.13 Function Calls

A function is a name for a particular calculation. This enables you to ask for it by name at any point in the program. For example, the function sqrt computes the square root of a number.

A fixed set of functions are built-in, which means they are available in every awk program. The sqrt function is one of these. See Built-in, for a list of built-in functions and their descriptions. In addition, you can define functions for use in your program. See User-defined, for instructions on how to do this.

The way to use a function is with a function call expression, which consists of the function name followed immediately by a list of arguments in parentheses. The arguments are expressions that provide the raw materials for the function's calculations. When there is more than one argument, they are separated by commas. If there are no arguments, just write ‘()’ after the function name. The following examples show function calls with and without arguments:

     sqrt(x^2 + y^2)        one argument
     atan2(y, x)            two arguments
     rand()                 no arguments

Caution: Do not put any space between the function name and the open-parenthesis! A user-defined function name looks just like the name of a variable—a space would make the expression look like concatenation of a variable with an expression inside parentheses.

With built-in functions, space before the parenthesis is harmless, but it is best not to get into the habit of using space to avoid mistakes with user-defined functions. Each function expects a particular number of arguments. For example, the sqrt function must be called with a single argument, the number of which to take the square root:


Some of the built-in functions have one or more optional arguments. If those arguments are not supplied, the functions use a reasonable default value. See Built-in, for full details. If arguments are omitted in calls to user-defined functions, then those arguments are treated as local variables and initialized to the empty string (see User-defined).

Like every other expression, the function call has a value, which is computed by the function based on the arguments you give it. In this example, the value of ‘sqrt(argument)’ is the square root of argument. The following program reads numbers, one number per line, and prints the square root of each one:

     $ awk '{ print "The square root of", $1, "is", sqrt($1) }'
     -| The square root of 1 is 1
     -| The square root of 3 is 1.73205
     -| The square root of 5 is 2.23607

A function can also have side effects, such as assigning values to certain variables or doing I/O. This program shows how the ‘match’ function (see String Functions) changes the variables RSTART and RLENGTH:

         if (match($1, $2))
             print RSTART, RLENGTH
             print "no match"

Here is a sample run:

     $ awk -f matchit.awk
     aaccdd  c+
     -| 3 2
     foo     bar
     -| no match
     abcdefg e
     -| 5 1

Previous: Function Calls, Up: Expressions

5.14 Operator Precedence (How Operators Nest)

Operator precedence determines how operators are grouped when different operators appear close by in one expression. For example, ‘*’ has higher precedence than ‘+’; thus, ‘a + b * c’ means to multiply b and c, and then add a to the product (i.e., ‘a + (b * c)’).

The normal precedence of the operators can be overruled by using parentheses. Think of the precedence rules as saying where the parentheses are assumed to be. In fact, it is wise to always use parentheses whenever there is an unusual combination of operators, because other people who read the program may not remember what the precedence is in this case. Even experienced programmers occasionally forget the exact rules, which leads to mistakes. Explicit parentheses help prevent any such mistakes.

When operators of equal precedence are used together, the leftmost operator groups first, except for the assignment, conditional, and exponentiation operators, which group in the opposite order. Thus, ‘a - b + c’ groups as ‘(a - b) + c’ and ‘a = b = c’ groups as ‘a = (b = c)’.

Normally the precedence of prefix unary operators does not matter, because there is only one way to interpret them: innermost first. Thus, ‘$++i’ means ‘$(++i)’ and ‘++$x’ means ‘++($x)’. However, when another operator follows the operand, then the precedence of the unary operators can matter. ‘$x^2’ means ‘($x)^2’, but ‘-x^2’ means ‘-(x^2)’, because ‘-’ has lower precedence than ‘^’, whereas ‘$’ has higher precedence. Also, operators cannot be combined in a way that violates the precedence rules; for example, ‘$$0++--’ is not a valid expression because the first ‘$’ has higher precedence than the ‘++’; to avoid the problem the expression can be rewritten as ‘$($0++)--’.

This table presents awk's operators, in order of highest to lowest precedence:



++ --
Increment, decrement.

^ **
Exponentiation. These operators group right-to-left.

+ - !
Unary plus, minus, logical “not.”

* / %
Multiplication, division, remainder.

+ -
Addition, subtraction.
String Concatenation
No special symbol is used to indicate concatenation. The operands are simply written side by side (see Concatenation).

< <= == !=
> >= >> | |&
Relational and redirection. The relational operators and the redirections have the same precedence level. Characters such as ‘>’ serve both as relationals and as redirections; the context distinguishes between the two meanings.

Note that the I/O redirection operators in print and printf statements belong to the statement level, not to expressions. The redirection does not produce an expression that could be the operand of another operator. As a result, it does not make sense to use a redirection operator near another operator of lower precedence without parentheses. Such combinations (for example, ‘print foo > a ? b : c’), result in syntax errors. The correct way to write this statement is ‘print foo > (a ? b : c)’.

~ !~
Matching, nonmatching.

Array membership.

Logical “and”.

Logical “or”.

Conditional. This operator groups right-to-left.

= += -= *=
/= %= ^= **=
Assignment. These operators group right to left.

NOTE: The ‘|&’, ‘**’, and ‘**=’ operators are not specified by POSIX. For maximum portability, do not use them.

Next: , Previous: Expressions, Up: Top

6 Patterns, Actions, and Variables

As you have already seen, each awk statement consists of a pattern with an associated action. This chapter describes how you build patterns and actions, what kinds of things you can do within actions, and awk's built-in variables.

The pattern-action rules and the statements available for use within actions form the core of awk programming. In a sense, everything covered up to here has been the foundation that programs are built on top of. Now it's time to start building something useful.

Next: , Up: Patterns and Actions

6.1 Pattern Elements

Patterns in awk control the execution of rules—a rule is executed when its pattern matches the current input record. The following is a summary of the types of awk patterns:

/regular expression/
A regular expression. It matches when the text of the input record fits the regular expression. (See Regexp.)
A single expression. It matches when its value is nonzero (if a number) or non-null (if a string). (See Expression Patterns.)
pat1, pat2
A pair of patterns separated by a comma, specifying a range of records. The range includes both the initial record that matches pat1 and the final record that matches pat2. (See Ranges.)
Special patterns for you to supply startup or cleanup actions for your awk program. (See BEGIN/END.)
The empty pattern matches every input record. (See Empty.)

Next: , Up: Pattern Overview

6.1.1 Regular Expressions as Patterns

Regular expressions are one of the first kinds of patterns presented in this book. This kind of pattern is simply a regexp constant in the pattern part of a rule. Its meaning is ‘$0 ~ /pattern/’. The pattern matches when the input record matches the regexp. For example:

     /foo|bar|baz/  { buzzwords++ }
     END            { print buzzwords, "buzzwords seen" }

Next: , Previous: Regexp Patterns, Up: Pattern Overview

6.1.2 Expressions as Patterns

Any awk expression is valid as an awk pattern. The pattern matches if the expression's value is nonzero (if a number) or non-null (if a string). The expression is reevaluated each time the rule is tested against a new input record. If the expression uses fields such as $1, the value depends directly on the new input record's text; otherwise, it depends on only what has happened so far in the execution of the awk program.

Comparison expressions, using the comparison operators described in Typing and Comparison, are a very common kind of pattern. Regexp matching and nonmatching are also very common expressions. The left operand of the ‘~’ and ‘!~’ operators is a string. The right operand is either a constant regular expression enclosed in slashes (/regexp/), or any expression whose string value is used as a dynamic regular expression (see Computed Regexps). The following example prints the second field of each input record whose first field is precisely ‘foo’:

     $ awk '$1 == "foo" { print $2 }' BBS-list

(There is no output, because there is no BBS site with the exact name ‘foo’.) Contrast this with the following regular expression match, which accepts any record with a first field that contains ‘foo’:

     $ awk '$1 ~ /foo/ { print $2 }' BBS-list
     -| 555-1234
     -| 555-6699
     -| 555-6480
     -| 555-2127

A regexp constant as a pattern is also a special case of an expression pattern. The expression /foo/ has the value one if ‘foo’ appears in the current input record. Thus, as a pattern, /foo/ matches any record containing ‘foo’.

Boolean expressions are also commonly used as patterns. Whether the pattern matches an input record depends on whether its subexpressions match. For example, the following command prints all the records in BBS-list that contain both ‘2400’ and ‘foo’:

     $ awk '/2400/ && /foo/' BBS-list
     -| fooey        555-1234     2400/1200/300     B

The following command prints all records in BBS-list that contain either2400’ or ‘foo’ (or both, of course):

     $ awk '/2400/ || /foo/' BBS-list
     -| alpo-net     555-3412     2400/1200/300     A
     -| bites        555-1675     2400/1200/300     A
     -| fooey        555-1234     2400/1200/300     B
     -| foot         555-6699     1200/300          B
     -| macfoo       555-6480     1200/300          A
     -| sdace        555-3430     2400/1200/300     A
     -| sabafoo      555-2127     1200/300          C

The following command prints all records in BBS-list that do not contain the string ‘foo’:

     $ awk '! /foo/' BBS-list
     -| aardvark     555-5553     1200/300          B
     -| alpo-net     555-3412     2400/1200/300     A
     -| barfly       555-7685     1200/300          A
     -| bites        555-1675     2400/1200/300     A
     -| camelot      555-0542     300               C
     -| core         555-2912     1200/300          C
     -| sdace        555-3430     2400/1200/300     A

The subexpressions of a Boolean operator in a pattern can be constant regular expressions, comparisons, or any other awk expressions. Range patterns are not expressions, so they cannot appear inside Boolean patterns. Likewise, the special patterns BEGIN and END, which never match any input record, are not expressions and cannot appear inside Boolean patterns.

Next: , Previous: Expression Patterns, Up: Pattern Overview

6.1.3 Specifying Record Ranges with Patterns

A range pattern is made of two patterns separated by a comma, in the form ‘begpat, endpat’. It is used to match ranges of consecutive input records. The first pattern, begpat, controls where the range begins, while endpat controls where the pattern ends. For example, the following:

     awk '$1 == "on", $1 == "off"' myfile

prints every record in myfile between ‘on’/‘off’ pairs, inclusive.

A range pattern starts out by matching begpat against every input record. When a record matches begpat, the range pattern is turned on and the range pattern matches this record as well. As long as the range pattern stays turned on, it automatically matches every input record read. The range pattern also matches endpat against every input record; when this succeeds, the range pattern is turned off again for the following record. Then the range pattern goes back to checking begpat against each record.

The record that turns on the range pattern and the one that turns it off both match the range pattern. If you don't want to operate on these records, you can write if statements in the rule's action to distinguish them from the records you are interested in.

It is possible for a pattern to be turned on and off by the same record. If the record satisfies both conditions, then the action is executed for just that record. For example, suppose there is text between two identical markers (e.g., the ‘%’ symbol), each on its own line, that should be ignored. A first attempt would be to combine a range pattern that describes the delimited text with the next statement (not discussed yet, see Next Statement). This causes awk to skip any further processing of the current record and start over again with the next input record. Such a program looks like this:

     /^%$/,/^%$/    { next }
                    { print }

This program fails because the range pattern is both turned on and turned off by the first line, which just has a ‘%’ on it. To accomplish this task, write the program in the following manner, using a flag:

     /^%$/     { skip = ! skip; next }
     skip == 1 { next } # skip lines with `skip' set

In a range pattern, the comma (‘,’) has the lowest precedence of all the operators (i.e., it is evaluated last). Thus, the following program attempts to combine a range pattern with another, simpler test:

     echo Yes | awk '/1/,/2/ || /Yes/'

The intent of this program is ‘(/1/,/2/) || /Yes/’. However, awk interprets this as ‘/1/, (/2/ || /Yes/)’. This cannot be changed or worked around; range patterns do not combine with other patterns:

     $ echo Yes | gawk '(/1/,/2/) || /Yes/'
     error--> gawk: cmd. line:1: (/1/,/2/) || /Yes/
     error--> gawk: cmd. line:1:           ^ parse error
     error--> gawk: cmd. line:2: (/1/,/2/) || /Yes/
     error--> gawk: cmd. line:2:                   ^ unexpected newline

Next: , Previous: Ranges, Up: Pattern Overview

6.1.4 The BEGIN and END Special Patterns

All the patterns described so far are for matching input records. The BEGIN and END special patterns are different. They supply startup and cleanup actions for awk programs. BEGIN and END rules must have actions; there is no default action for these rules because there is no current record when they run. BEGIN and END rules are often referred to as “BEGIN and END blocks” by long-time awk programmers.

Next: , Up: BEGIN/END Startup and Cleanup Actions

A BEGIN rule is executed once only, before the first input record is read. Likewise, an END rule is executed once only, after all the input is read. For example:

     $ awk '
     > BEGIN { print "Analysis of \"foo\"" }
     > /foo/ { ++n }
     > END   { print "\"foo\" appears", n, "times." }' BBS-list
     -| Analysis of "foo"
     -| "foo" appears 4 times.

This program finds the number of records in the input file BBS-list that contain the string ‘foo’. The BEGIN rule prints a title for the report. There is no need to use the BEGIN rule to initialize the counter n to zero, since awk does this automatically (see Variables). The second rule increments the variable n every time a record containing the pattern ‘foo’ is read. The END rule prints the value of n at the end of the run.

The special patterns BEGIN and END cannot be used in ranges or with Boolean operators (indeed, they cannot be used with any operators). An awk program may have multiple BEGIN and/or END rules. They are executed in the order in which they appear: all the BEGIN rules at startup and all the END rules at termination. BEGIN and END rules may be intermixed with other rules. This feature was added in the 1987 version of awk and is included in the POSIX standard. The original (1978) version of awk required the BEGIN rule to be placed at the beginning of the program, the END rule to be placed at the end, and only allowed one of each. This is no longer required, but it is a good idea to follow this template in terms of program organization and readability.

Multiple BEGIN and END rules are useful for writing library functions, because each library file can have its own BEGIN and/or END rule to do its own initialization and/or cleanup. The order in which library functions are named on the command line controls the order in which their BEGIN and END rules are executed. Therefore, you have to be careful when writing such rules in library files so that the order in which they are executed doesn't matter. See Options, for more information on using library functions. See Library Functions, for a number of useful library functions.

If an awk program has only a BEGIN rule and no other rules, then the program exits after the BEGIN rule is run.28 However, if an END rule exists, then the input is read, even if there are no other rules in the program. This is necessary in case the END rule checks the FNR and NR variables.

Previous: Using BEGIN/END, Up: BEGIN/END Input/Output from BEGIN and END Rules

There are several (sometimes subtle) points to remember when doing I/O from a BEGIN or END rule. The first has to do with the value of $0 in a BEGIN rule. Because BEGIN rules are executed before any input is read, there simply is no input record, and therefore no fields, when executing BEGIN rules. References to $0 and the fields yield a null string or zero, depending upon the context. One way to give $0 a real value is to execute a getline command without a variable (see Getline). Another way is simply to assign a value to $0.

The second point is similar to the first but from the other direction. Traditionally, due largely to implementation issues, $0 and NF were undefined inside an END rule. The POSIX standard specifies that NF is available in an END rule. It contains the number of fields from the last input record. Most probably due to an oversight, the standard does not say that $0 is also preserved, although logically one would think that it should be. In fact, gawk does preserve the value of $0 for use in END rules. Be aware, however, that Unix awk, and possibly other implementations, do not.

The third point follows from the first two. The meaning of ‘print’ inside a BEGIN or END rule is the same as always: ‘print $0’. If $0 is the null string, then this prints an empty line. Many long time awk programmers use an unadorned ‘print’ in BEGIN and END rules, to mean ‘print ""’, relying on $0 being null. Although one might generally get away with this in BEGIN rules, it is a very bad idea in END rules, at least in gawk. It is also poor style, since if an empty line is needed in the output, the program should print one explicitly.

Finally, the next and nextfile statements are not allowed in a BEGIN rule, because the implicit read-a-record-and-match-against-the-rules loop has not started yet. Similarly, those statements are not valid in an END rule, since all the input has been read. (See Next Statement, and see Nextfile Statement.)

Previous: BEGIN/END, Up: Pattern Overview

6.1.5 The Empty Pattern

An empty (i.e., nonexistent) pattern is considered to match every input record. For example, the program:

     awk '{ print $1 }' BBS-list

prints the first field of every record.

Next: , Previous: Pattern Overview, Up: Patterns and Actions

6.2 Using Shell Variables in Programs

awk programs are often used as components in larger programs written in shell. For example, it is very common to use a shell variable to hold a pattern that the awk program searches for. There are two ways to get the value of the shell variable into the body of the awk program.

The most common method is to use shell quoting to substitute the variable's value into the program inside the script. For example, in the following program:

     echo -n "Enter search pattern: "
     read pattern
     awk "/$pattern/ "'{ nmatches++ }
          END { print nmatches, "found" }' /path/to/data

the awk program consists of two pieces of quoted text that are concatenated together to form the program. The first part is double-quoted, which allows substitution of the pattern variable inside the quotes. The second part is single-quoted.

Variable substitution via quoting works, but can be potentially messy. It requires a good understanding of the shell's quoting rules (see Quoting), and it's often difficult to correctly match up the quotes when reading the program.

A better method is to use awk's variable assignment feature (see Assignment Options) to assign the shell variable's value to an awk variable's value. Then use dynamic regexps to match the pattern (see Computed Regexps). The following shows how to redo the previous example using this technique:

     echo -n "Enter search pattern: "
     read pattern
     awk -v pat="$pattern" '$0 ~ pat { nmatches++ }
            END { print nmatches, "found" }' /path/to/data

Now, the awk program is just one single-quoted string. The assignment ‘-v pat="$pattern"’ still requires double quotes, in case there is whitespace in the value of $pattern. The awk variable pat could be named pattern too, but that would be more confusing. Using a variable also provides more flexibility, since the variable can be used anywhere inside the program—for printing, as an array subscript, or for any other use—without requiring the quoting tricks at every point in the program.

Next: , Previous: Using Shell Variables, Up: Patterns and Actions

6.3 Actions

An awk program or script consists of a series of rules and function definitions interspersed. (Functions are described later. See User-defined.) A rule contains a pattern and an action, either of which (but not both) may be omitted. The purpose of the action is to tell awk what to do once a match for the pattern is found. Thus, in outline, an awk program generally looks like this:

     [pattern] [{ action }]
     [pattern] [{ action }]
     function name(args) { ... }

An action consists of one or more awk statements, enclosed in curly braces (‘{...}’). Each statement specifies one thing to do. The statements are separated by newlines or semicolons. The curly braces around an action must be used even if the action contains only one statement, or if it contains no statements at all. However, if you omit the action entirely, omit the curly braces as well. An omitted action is equivalent to ‘{ print $0 }’:

     /foo/  { }     match foo, do nothing --- empty action
     /foo/          match foo, print the record --- omitted action

The following types of statements are supported in awk:

Call functions or assign values to variables (see Expressions). Executing this kind of statement simply computes the value of the expression. This is useful when the expression has side effects (see Assignment Ops).
Control statements
Specify the control flow of awk programs. The awk language gives you C-like constructs (if, for, while, and do) as well as a few special ones (see Statements).
Compound statements
Consist of one or more statements enclosed in curly braces. A compound statement is used in order to put several statements together in the body of an if, while, do, or for statement.
Input statements
Use the getline command (see Getline). Also supplied in awk are the next statement (see Next Statement), and the nextfile statement (see Nextfile Statement).
Output statements
Such as print and printf. See Printing.
Deletion statements
For deleting array elements. See Delete.

Next: , Previous: Action Overview, Up: Patterns and Actions

6.4 Control Statements in Actions

Control statements, such as if, while, and so on, control the flow of execution in awk programs. Most of the control statements in awk are patterned on similar statements in C.

All the control statements start with special keywords, such as if and while, to distinguish them from simple expressions. Many control statements contain other statements. For example, the if statement contains another statement that may or may not be executed. The contained statement is called the body. To include more than one statement in the body, group them into a single compound statement with curly braces, separating them with newlines or semicolons.

Next: , Up: Statements

6.4.1 The if-else Statement

The if-else statement is awk's decision-making statement. It looks like this:

     if (condition) then-body [else else-body]

The condition is an expression that controls what the rest of the statement does. If the condition is true, then-body is executed; otherwise, else-body is executed. The else part of the statement is optional. The condition is considered false if its value is zero or the null string; otherwise, the condition is true. Refer to the following:

     if (x % 2 == 0)
         print "x is even"
         print "x is odd"

In this example, if the expression ‘x % 2 == 0’ is true (that is, if the value of x is evenly divisible by two), then the first print statement is executed; otherwise, the second print statement is executed. If the else keyword appears on the same line as then-body and then-body is not a compound statement (i.e., not surrounded by curly braces), then a semicolon must separate then-body from the else. To illustrate this, the previous example can be rewritten as:

     if (x % 2 == 0) print "x is even"; else
             print "x is odd"

If the ‘;’ is left out, awk can't interpret the statement and it produces a syntax error. Don't actually write programs this way, because a human reader might fail to see the else if it is not the first thing on its line.

Next: , Previous: If Statement, Up: Statements

6.4.2 The while Statement

In programming, a loop is a part of a program that can be executed two or more times in succession. The while statement is the simplest looping statement in awk. It repeatedly executes a statement as long as a condition is true. For example:

     while (condition)

body is a statement called the body of the loop, and condition is an expression that controls how long the loop keeps running. The first thing the while statement does is test the condition. If the condition is true, it executes the statement body. After body has been executed, condition is tested again, and if it is still true, body is executed again. This process repeats until the condition is no longer true. If the condition is initially false, the body of the loop is never executed and awk continues with the statement following the loop. This example prints the first three fields of each record, one per line:

     awk '{ i = 1
            while (i <= 3) {
                print $i
     }' inventory-shipped

The body of this loop is a compound statement enclosed in braces, containing two statements. The loop works in the following manner: first, the value of i is set to one. Then, the while statement tests whether i is less than or equal to three. This is true when i equals one, so the i-th field is printed. Then the ‘i++’ increments the value of i and the loop repeats. The loop terminates when i reaches four.

A newline is not required between the condition and the body; however using one makes the program clearer unless the body is a compound statement or else is very simple. The newline after the open-brace that begins the compound statement is not required either, but the program is harder to read without it.

Next: , Previous: While Statement, Up: Statements

6.4.3 The do-while Statement

The do loop is a variation of the while looping statement. The do loop executes the body once and then repeats the body as long as the condition is true. It looks like this:

     while (condition)

Even if the condition is false at the start, the body is executed at least once (and only once, unless executing body makes condition true). Contrast this with the corresponding while statement:

     while (condition)

This statement does not execute body even once if the condition is false to begin with. The following is an example of a do statement:

     {      i = 1
            do {
               print $0
            } while (i <= 10)

This program prints each input record 10 times. However, it isn't a very realistic example, since in this case an ordinary while would do just as well. This situation reflects actual experience; only occasionally is there a real use for a do statement.

Next: , Previous: Do Statement, Up: Statements

6.4.4 The for Statement

The for statement makes it more convenient to count iterations of a loop. The general form of the for statement looks like this:

     for (initialization; condition; increment)

The initialization, condition, and increment parts are arbitrary awk expressions, and body stands for any awk statement.

The for statement starts by executing initialization. Then, as long as the condition is true, it repeatedly executes body and then increment. Typically, initialization sets a variable to either zero or one, increment adds one to it, and condition compares it against the desired number of iterations. For example:

     awk '{ for (i = 1; i <= 3; i++)
               print $i
     }' inventory-shipped

This prints the first three fields of each input record, with one field per line.

It isn't possible to set more than one variable in the initialization part without using a multiple assignment statement such as ‘x = y = 0’. This makes sense only if all the initial values are equal. (But it is possible to initialize additional variables by writing their assignments as separate statements preceding the for loop.)

The same is true of the increment part. Incrementing additional variables requires separate statements at the end of the loop. The C compound expression, using C's comma operator, is useful in this context but it is not supported in awk.

Most often, increment is an increment expression, as in the previous example. But this is not required; it can be any expression whatsoever. For example, the following statement prints all the powers of two between 1 and 100:

     for (i = 1; i <= 100; i *= 2)
       print i

If there is nothing to be done, any of the three expressions in the parentheses following the for keyword may be omitted. Thus, ‘for (; x > 0;) is equivalent to ‘while (x > 0). If the condition is omitted, it is treated as true, effectively yielding an infinite loop (i.e., a loop that never terminates).

In most cases, a for loop is an abbreviation for a while loop, as shown here:

     while (condition) {

The only exception is when the continue statement (see Continue Statement) is used inside the loop. Changing a for statement to a while statement in this way can change the effect of the continue statement inside the loop.

The awk language has a for statement in addition to a while statement because a for loop is often both less work to type and more natural to think of. Counting the number of iterations is very common in loops. It can be easier to think of this counting as part of looping rather than as something to do inside the loop.

Next: , Previous: For Statement, Up: Statements

6.4.5 The switch Statement

NOTE: This subsection describes an experimental feature added in gawk 3.1.3. It is not enabled by default. To enable it, use the --enable-switch option to configure when gawk is being configured and built. See Additional Configuration Options, for more information.

The switch statement allows the evaluation of an expression and the execution of statements based on a case match. Case statements are checked for a match in the order they are defined. If no suitable case is found, the default section is executed, if supplied.

Each case contains a single constant, be it numeric, string, or regexp. The switch expression is evaluated, and then each case's constant is compared against the result in turn. The type of constant determines the comparison: numeric or string do the usual comparisons. A regexp constant does a regular expression match against the string value of the original expression. The general form of the switch statement looks like this:

     switch (expression) {
     case value or regular expression:

Control flow in the switch statement works as it does in C. Once a match to a given case is made, case statement bodies are executed until a break, continue, next, nextfile or exit is encountered, or the end of the switch statement itself. For example:

     switch (NR * 2 + 1) {
     case 3:
     case "11":
         print NR - 1
     case /2[[:digit:]]+/:
         print NR
         print NR + 1
     case -1:
         print NR * -1

Note that if none of the statements specified above halt execution of a matched case statement, execution falls through to the next case until execution halts. In the above example, for any case value starting with ‘2’ followed by one or more digits, the print statement is executed and then falls through into the default section, executing its print statement. In turn, the −1 case will also be executed since the default does not halt execution.

Next: , Previous: Switch Statement, Up: Statements

6.4.6 The break Statement

The break statement jumps out of the innermost for, while, or do loop that encloses it. The following example finds the smallest divisor of any integer, and also identifies prime numbers:

     # find smallest divisor of num
        num = $1
        for (div = 2; div*div <= num; div++)
          if (num % div == 0)
        if (num % div == 0)
          printf "Smallest divisor of %d is %d\n", num, div
          printf "%d is prime\n", num

When the remainder is zero in the first if statement, awk immediately breaks out of the containing for loop. This means that awk proceeds immediately to the statement following the loop and continues processing. (This is very different from the exit statement, which stops the entire awk program. See Exit Statement.)

Th following program illustrates how the condition of a for or while statement could be replaced with a break inside an if:

     # find smallest divisor of num
       num = $1
       for (div = 2; ; div++) {
         if (num % div == 0) {
           printf "Smallest divisor of %d is %d\n", num, div
         if (div*div > num) {
           printf "%d is prime\n", num

The break statement has no meaning when used outside the body of a loop. However, although it was never documented, historical implementations of awk treated the break statement outside of a loop as if it were a next statement (see Next Statement). Recent versions of Unix awk no longer allow this usage. gawk supports this use of break only if --traditional has been specified on the command line (see Options). Otherwise, it is treated as an error, since the POSIX standard specifies that break should only be used inside the body of a loop. (d.c.)

Next: , Previous: Break Statement, Up: Statements

6.4.7 The continue Statement

As with break, the continue statement is used only inside for, while, and do loops. It skips over the rest of the loop body, causing the next cycle around the loop to begin immediately. Contrast this with break, which jumps out of the loop altogether.

The continue statement in a for loop directs awk to skip the rest of the body of the loop and resume execution with the increment-expression of the for statement. The following program illustrates this fact:

     BEGIN {
          for (x = 0; x <= 20; x++) {
              if (x == 5)
              printf "%d ", x
          print ""

This program prints all the numbers from 0 to 20—except for 5, for which the printf is skipped. Because the increment ‘x++’ is not skipped, x does not remain stuck at 5. Contrast the for loop from the previous example with the following while loop:

     BEGIN {
          x = 0
          while (x <= 20) {
              if (x == 5)
              printf "%d ", x
          print ""

This program loops forever once x reaches 5.

The continue statement has no meaning when used outside the body of a loop. Historical versions of awk treated a continue statement outside a loop the same way they treated a break statement outside a loop: as if it were a next statement (see Next Statement). Recent versions of Unix awk no longer work this way, and gawk allows it only if --traditional is specified on the command line (see Options). Just like the break statement, the POSIX standard specifies that continue should only be used inside the body of a loop. (d.c.)

Next: , Previous: Continue Statement, Up: Statements

6.4.8 The next Statement

The next statement forces awk to immediately stop processing the current record and go on to the next record. This means that no further rules are executed for the current record, and the rest of the current rule's action isn't executed.

Contrast this with the effect of the getline function (see Getline). That also causes awk to read the next record immediately, but it does not alter the flow of control in any way (i.e., the rest of the current action executes with a new input record).

At the highest level, awk program execution is a loop that reads an input record and then tests each rule's pattern against it. If you think of this loop as a for statement whose body contains the rules, then the next statement is analogous to a continue statement. It skips to the end of the body of this implicit loop and executes the increment (which reads another record).

For example, suppose an awk program works only on records with four fields, and it shouldn't fail when given bad input. To avoid complicating the rest of the program, write a “weed out” rule near the beginning, in the following manner:

     NF != 4 {
       err = sprintf("%s:%d: skipped: NF != 4\n", FILENAME, FNR)
       print err > "/dev/stderr"

Because of the next statement, the program's subsequent rules won't see the bad record. The error message is redirected to the standard error output stream, as error messages should be. For more detail see Special Files.

According to the POSIX standard, the behavior is undefined if the next statement is used in a BEGIN or END rule. gawk treats it as a syntax error. Although POSIX permits it, some other awk implementations don't allow the next statement inside function bodies (see User-defined). Just as with any other next statement, a next statement inside a function body reads the next record and starts processing it with the first rule in the program. If the next statement causes the end of the input to be reached, then the code in any END rules is executed. See BEGIN/END.

Next: , Previous: Next Statement, Up: Statements

6.4.9 Using gawk's nextfile Statement

gawk provides the nextfile statement, which is similar to the next statement. However, instead of abandoning processing of the current record, the nextfile statement instructs gawk to stop processing the current data file.

The nextfile statement is a gawk extension. In most other awk implementations, or if gawk is in compatibility mode (see Options), nextfile is not special.

Upon execution of the nextfile statement, FILENAME is updated to the name of the next data file listed on the command line, FNR is reset to one, ARGIND is incremented, and processing starts over with the first rule in the program. (ARGIND hasn't been introduced yet. See Built-in Variables.) If the nextfile statement causes the end of the input to be reached, then the code in any END rules is executed. See BEGIN/END.

The nextfile statement is useful when there are many data files to process but it isn't necessary to process every record in every file. Normally, in order to move on to the next data file, a program has to continue scanning the unwanted records. The nextfile statement accomplishes this much more efficiently.

While one might think that ‘close(FILENAME)’ would accomplish the same as nextfile, this isn't true. close is reserved for closing files, pipes, and coprocesses that are opened with redirections. It is not related to the main processing that awk does with the files listed in ARGV.

If it's necessary to use an awk version that doesn't support nextfile, see Nextfile Function, for a user-defined function that simulates the nextfile statement.

The current version of the Bell Laboratories awk (see Other Versions) also supports nextfile. However, it doesn't allow the nextfile statement inside function bodies (see User-defined). gawk does; a nextfile inside a function body reads the next record and starts processing it with the first rule in the program, just as any other nextfile statement.

Caution: Versions of gawk prior to 3.0 used two words (‘next file’) for the nextfile statement. In version 3.0, this was changed to one word, because the treatment of ‘file’ was inconsistent. When it appeared after next, ‘file’ was a keyword; otherwise, it was a regular identifier. The old usage is no longer accepted; ‘next file’ generates a syntax error.

Previous: Nextfile Statement, Up: Statements

6.4.10 The exit Statement

The exit statement causes awk to immediately stop executing the current rule and to stop processing input; any remaining input is ignored. The exit statement is written as follows:

     exit [return code]

When an exit statement is executed from a BEGIN rule, the program stops processing everything immediately. No input records are read. However, if an END rule is present, as part of executing the exit statement, the END rule is executed (see BEGIN/END). If exit is used as part of an END rule, it causes the program to stop immediately.

An exit statement that is not part of a BEGIN or END rule stops the execution of any further automatic rules for the current record, skips reading any remaining input records, and executes the END rule if there is one.

In such a case, if you don't want the END rule to do its job, set a variable to nonzero before the exit statement and check that variable in the END rule. See Assert Function, for an example that does this.

If an argument is supplied to exit, its value is used as the exit status code for the awk process. If no argument is supplied, exit returns status zero (success). In the case where an argument is supplied to a first exit statement, and then exit is called a second time from an END rule with no argument, awk uses the previously supplied exit value. (d.c.)

For example, suppose an error condition occurs that is difficult or impossible to handle. Conventionally, programs report this by exiting with a nonzero status. An awk program can do this using an exit statement with a nonzero argument, as shown in the following example:

     BEGIN {
            if (("date" | getline date_now) <= 0) {
              print "Can't get system date" > "/dev/stderr"
              exit 1
            print "current date is", date_now

Previous: Statements, Up: Patterns and Actions

6.5 Built-in Variables

Most awk variables are available to use for your own purposes; they never change unless your program assigns values to them, and they never affect anything unless your program examines them. However, a few variables in awk have special built-in meanings. awk examines some of these automatically, so that they enable you to tell awk how to do certain things. Others are set automatically by awk, so that they carry information from the internal workings of awk to your program.

This section documents all the built-in variables of gawk, most of which are also documented in the chapters describing their areas of activity.

Next: , Up: Built-in Variables

6.5.1 Built-in Variables That Control awk

The following is an alphabetical list of variables that you can change to control how awk does certain things. The variables that are specific to gawk are marked with a pound sign (‘#’).

On non-POSIX systems, this variable specifies use of binary mode for all I/O. Numeric values of one, two, or three specify that input files, output files, or all files, respectively, should use binary I/O. Alternatively, string values of "r" or "w" specify that input files and output files, respectively, should use binary I/O. A string value of "rw" or "wr" indicates that all files should use binary I/O. Any other string value is equivalent to "rw", but gawk generates a warning message. BINMODE is described in more detail in PC Using.

This variable is a gawk extension. In other awk implementations (except mawk, see Other Versions), or if gawk is in compatibility mode (see Options), it is not special.

This string controls conversion of numbers to strings (see Conversion). It works by being passed, in effect, as the first argument to the sprintf function (see String Functions). Its default value is "%.6g". CONVFMT was introduced by the POSIX standard.

This is a space-separated list of columns that tells gawk how to split input with fixed columnar boundaries. Assigning a value to FIELDWIDTHS overrides the use of FS for field splitting. See Constant Size, for more information.

If gawk is in compatibility mode (see Options), then FIELDWIDTHS has no special meaning, and field-splitting operations occur based exclusively on the value of FS.

This is the input field separator (see Field Separators). The value is a single-character string or a multi-character regular expression that matches the separations between fields in an input record. If the value is the null string (""), then each character in the record becomes a separate field. (This behavior is a gawk extension. POSIX awk does not specify the behavior when FS is the null string.)

The default value is " ", a string consisting of a single space. As a special exception, this value means that any sequence of spaces, tabs, and/or newlines is a single separator.29 It also causes spaces, tabs, and newlines at the beginning and end of a record to be ignored.

You can set the value of FS on the command line using the -F option:

          awk -F, 'program' input-files

If gawk is using FIELDWIDTHS for field splitting, assigning a value to FS causes gawk to return to the normal, FS-based field splitting. An easy way to do this is to simply say ‘FS = FS’, perhaps with an explanatory comment.

If IGNORECASE is nonzero or non-null, then all string comparisons and all regular expression matching are case independent. Thus, regexp matching with ‘~’ and ‘!~’, as well as the gensub, gsub, index, match, split, and sub functions, record termination with RS, and field splitting with FS, all ignore case when doing their particular regexp operations. However, the value of IGNORECASE does not affect array subscripting and it does not affect field splitting when using a single-character field separator. See Case-sensitivity.

If gawk is in compatibility mode (see Options), then IGNORECASE has no special meaning. Thus, string and regexp operations are always case-sensitive.

When this variable is true (nonzero or non-null), gawk behaves as if the --lint command-line option is in effect. (see Options). With a value of "fatal", lint warnings become fatal errors. With a value of "invalid", only warnings about things that are actually invalid are issued. (This is not fully implemented yet.) Any other true value prints nonfatal warnings. Assigning a false value to LINT turns off the lint warnings.

This variable is a gawk extension. It is not special in other awk implementations. Unlike the other special variables, changing LINT does affect the production of lint warnings, even if gawk is in compatibility mode. Much as the --lint and --traditional options independently control different aspects of gawk's behavior, the control of lint warnings during program execution is independent of the flavor of awk being executed.

This string controls conversion of numbers to strings (see Conversion) for printing with the print statement. It works by being passed as the first argument to the sprintf function (see String Functions). Its default value is "%.6g". Earlier versions of awk also used OFMT to specify the format for converting numbers to strings in general expressions; this is now done by CONVFMT.

This is the output field separator (see Output Separators). It is output between the fields printed by a print statement. Its default value is " ", a string consisting of a single space.

This is the output record separator. It is output at the end of every print statement. Its default value is "\n", the newline character. (See Output Separators.)

This is awk's input record separator. Its default value is a string containing a single newline character, which means that an input record consists of a single line of text. It can also be the null string, in which case records are separated by runs of blank lines. If it is a regexp, records are separated by matches of the regexp in the input text. (See Records.)

The ability for RS to be a regular expression is a gawk extension. In most other awk implementations, or if gawk is in compatibility mode (see Options), just the first character of RS's value is used.

This is the subscript separator. It has the default value of "\034" and is used to separate the parts of the indices of a multidimensional array. Thus, the expression foo["A", "B"] really accesses foo["A\034B"] (see Multi-dimensional).

This variable is used for internationalization of programs at the awk level. It sets the default text domain for specially marked string constants in the source text, as well as for the dcgettext, dcngettext and bindtextdomain functions (see Internationalization). The default value of TEXTDOMAIN is "messages".

This variable is a gawk extension. In other awk implementations, or if gawk is in compatibility mode (see Options), it is not special.

Next: , Previous: User-modified, Up: Built-in Variables

6.5.2 Built-in Variables That Convey Information

The following is an alphabetical list of variables that awk sets automatically on certain occasions in order to provide information to your program. The variables that are specific to gawk are marked with a pound sign (‘#’).

The command-line arguments available to awk programs are stored in an array called ARGV. ARGC is the number of command-line arguments present. See Other Arguments. Unlike most awk arrays, ARGV is indexed from 0 to ARGC − 1. In the following example:
          $ awk 'BEGIN {
          >         for (i = 0; i < ARGC; i++)
          >             print ARGV[i]
          >      }' inventory-shipped BBS-list
          -| awk
          -| inventory-shipped
          -| BBS-list

ARGV[0] contains "awk", ARGV[1] contains "inventory-shipped", and ARGV[2] contains "BBS-list". The value of ARGC is three, one more than the index of the last element in ARGV, because the elements are numbered from zero.

The names ARGC and ARGV, as well as the convention of indexing the array from 0 to ARGC − 1, are derived from the C language's method of accessing command-line arguments.

The value of ARGV[0] can vary from system to system. Also, you should note that the program text is not included in ARGV, nor are any of awk's command-line options. See ARGC and ARGV, for information about how awk uses these variables.

The index in ARGV of the current file being processed. Every time gawk opens a new data file for processing, it sets ARGIND to the index in ARGV of the file name. When gawk is processing the input files, ‘FILENAME == ARGV[ARGIND]’ is always true.

This variable is useful in file processing; it allows you to tell how far along you are in the list of data files as well as to distinguish between successive instances of the same file name on the command line.

While you can change the value of ARGIND within your awk program, gawk automatically sets it to a new value when the next file is opened.

This variable is a gawk extension. In other awk implementations, or if gawk is in compatibility mode (see Options), it is not special.

An associative array that contains the values of the environment. The array indices are the environment variable names; the elements are the values of the particular environment variables. For example, ENVIRON["HOME"] might be /home/arnold. Changing this array does not affect the environment passed on to any programs that awk may spawn via redirection or the system function.

Some operating systems may not have environment variables. On such systems, the ENVIRON array is empty (except for ENVIRON["AWKPATH"], see AWKPATH Variable).

If a system error occurs during a redirection for getline, during a read for getline, or during a close operation, then ERRNO contains a string describing the error.

ERRNO works similarly to the C variable errno. In particular gawk never clears it (sets it to zero or ""). Thus, you should only expect its value to be meaningful when an I/O operation returns a failure value, such as getline returning −1. You are, of course, free to clear it yourself before doing an I/O operation.

This variable is a gawk extension. In other awk implementations, or if gawk is in compatibility mode (see Options), it is not special.

The name of the file that awk is currently reading. When no data files are listed on the command line, awk reads from the standard input and FILENAME is set to "-". FILENAME is changed each time a new file is read (see Reading Files). Inside a BEGIN rule, the value of FILENAME is "", since there are no input files being processed yet.30 (d.c.) Note, though, that using getline (see Getline) inside a BEGIN rule can give FILENAME a value.

The current record number in the current file. FNR is incremented each time a new record is read (see Getline). It is reinitialized to zero each time a new input file is started.

The number of fields in the current input record. NF is set each time a new record is read, when a new field is created or when $0 changes (see Fields).

Unlike most of the variables described in this section, assigning a value to NF has the potential to affect awk's internal workings. In particular, assignments to NF can be used to create or remove fields from the current record: See Changing Fields.

The number of input records awk has processed since the beginning of the program's execution (see Records). NR is incremented each time a new record is read.

The elements of this array provide access to information about the running awk program. The following elements (listed alphabetically) are guaranteed to be available:
The value of the getegid system call.
The value of the geteuid system call.
This is "FS" if field splitting with FS is in effect, or it is "FIELDWIDTHS" if field splitting with FIELDWIDTHS is in effect.
The value of the getgid system call.
The process group ID of the current process.
The process ID of the current process.
The parent process ID of the current process.
The value of the getuid system call.
The version of gawk. This is available from version 3.1.4 and later.

On some systems, there may be elements in the array, "group1" through "groupN" for some N. N is the number of supplementary groups that the process has. Use the in operator to test for these elements (see Reference to Elements).

This array is a gawk extension. In other awk implementations, or if gawk is in compatibility mode (see Options), it is not special.

The length of the substring matched by the match function (see String Functions). RLENGTH is set by invoking the match function. Its value is the length of the matched string, or −1 if no match is found.

The start-index in characters of the substring that is matched by the match function (see String Functions). RSTART is set by invoking the match function. Its value is the position of the string where the matched substring starts, or zero if no match was found.

RT #
This is set each time a record is read. It contains the input text that matched the text denoted by RS, the record separator.

This variable is a gawk extension. In other awk implementations, or if gawk is in compatibility mode (see Options), it is not special.

Advanced Notes: Changing NR and FNR

awk increments NR and FNR each time it reads a record, instead of setting them to the absolute value of the number of records read. This means that a program can change these variables and their new values are incremented for each record. (d.c.) This is demonstrated in the following example:

     $ echo '1
     > 2
     > 3
     > 4' | awk 'NR == 2 { NR = 17 }
     > { print NR }'
     -| 1
     -| 17
     -| 18
     -| 19

Before FNR was added to the awk language (see V7/SVR3.1), many awk programs used this feature to track the number of records in a file by resetting NR to zero when FILENAME changed.

Previous: Auto-set, Up: Built-in Variables

6.5.3 Using ARGC and ARGV

Auto-set, presented the following program describing the information contained in ARGC and ARGV:

     $ awk 'BEGIN {
     >        for (i = 0; i < ARGC; i++)
     >            print ARGV[i]
     >      }' inventory-shipped BBS-list
     -| awk
     -| inventory-shipped
     -| BBS-list

In this example, ARGV[0] contains ‘awk’, ARGV[1] contains ‘inventory-shipped’, and ARGV[2] contains ‘BBS-list’. Notice that the awk program is not entered in ARGV. The other special command-line options, with their arguments, are also not entered. This includes variable assignments done with the -v option (see Options). Normal variable assignments on the command line are treated as arguments and do show up in the ARGV array:

     $ cat showargs.awk
     -| BEGIN {
     -|     printf "A=%d, B=%d\n", A, B
     -|     for (i = 0; i < ARGC; i++)
     -|         printf "\tARGV[%d] = %s\n", i, ARGV[i]
     -| }
     -| END   { printf "A=%d, B=%d\n", A, B }
     $ awk -v A=1 -f showargs.awk B=2 /dev/null
     -| A=1, B=0
     -|        ARGV[0] = awk
     -|        ARGV[1] = B=2
     -|        ARGV[2] = /dev/null
     -| A=1, B=2

A program can alter ARGC and the elements of ARGV. Each time awk reaches the end of an input file, it uses the next element of ARGV as the name of the next input file. By storing a different string there, a program can change which files are read. Use "-" to represent the standard input. Storing additional elements and incrementing ARGC causes additional files to be read.

If the value of ARGC is decreased, that eliminates input files from the end of the list. By recording the old value of ARGC elsewhere, a program can treat the eliminated arguments as something other than file names.

To eliminate a file from the middle of the list, store the null string ("") into ARGV in place of the file's name. As a special feature, awk ignores file names that have been replaced with the null string. Another option is to use the delete statement to remove elements from ARGV (see Delete).

All of these actions are typically done in the BEGIN rule, before actual processing of the input begins. See Split Program, and see Tee Program, for examples of each way of removing elements from ARGV. The following fragment processes ARGV in order to examine, and then remove, command-line options:

     BEGIN {
         for (i = 1; i < ARGC; i++) {
             if (ARGV[i] == "-v")
                 verbose = 1
             else if (ARGV[i] == "-d")
                 debug = 1
             else if (ARGV[i] ~ /^-./) {
                 e = sprintf("%s: unrecognized option -- %c",
                         ARGV[0], substr(ARGV[i], 2, 1))
                 print e > "/dev/stderr"
             } else
             delete ARGV[i]

To actually get the options into the awk program, end the awk options with -- and then supply the awk program's options, in the following manner:

     awk -f myprog -- -v -d file1 file2 ...

This is not necessary in gawk. Unless --posix has been specified, gawk silently puts any unrecognized options into ARGV for the awk program to deal with. As soon as it sees an unknown option, gawk stops looking for other options that it might otherwise recognize. The previous example with gawk would be:

     gawk -f myprog -d -v file1 file2 ...

Because -d is not a valid gawk option, it and the following -v are passed on to the awk program.

Next: , Previous: Patterns and Actions, Up: Top

7 Arrays in awk

An array is a table of values called elements. The elements of an array are distinguished by their indices. Indices may be either numbers or strings.

This chapter describes how arrays work in awk, how to use array elements, how to scan through every element in an array, and how to remove array elements. It also describes how awk simulates multidimensional arrays, as well as some of the less obvious points about array usage. The chapter finishes with a discussion of gawk's facility for sorting an array based on its indices.

awk maintains a single set of names that may be used for naming variables, arrays, and functions (see User-defined). Thus, you cannot have a variable and an array with the same name in the same awk program.

Next: , Up: Arrays

7.1 Introduction to Arrays

The awk language provides one-dimensional arrays for storing groups of related strings or numbers. Every awk array must have a name. Array names have the same syntax as variable names; any valid variable name would also be a valid array name. But one name cannot be used in both ways (as an array and as a variable) in the same awk program.

Arrays in awk superficially resemble arrays in other programming languages, but there are fundamental differences. In awk, it isn't necessary to specify the size of an array before starting to use it. Additionally, any number or string in awk, not just consecutive integers, may be used as an array index.

In most other languages, arrays must be declared before use, including a specification of how many elements or components they contain. In such languages, the declaration causes a contiguous block of memory to be allocated for that many elements. Usually, an index in the array must be a positive integer. For example, the index zero specifies the first element in the array, which is actually stored at the beginning of the block of memory. Index one specifies the second element, which is stored in memory right after the first element, and so on. It is impossible to add more elements to the array, because it has room only for as many elements as given in the declaration. (Some languages allow arbitrary starting and ending indices—e.g., ‘15 .. 27’—but the size of the array is still fixed when the array is declared.)

A contiguous array of four elements might look like the following example, conceptually, if the element values are 8, "foo", "", and 30:

Only the values are stored; the indices are implicit from the order of the values. Here, 8 is the value at index zero, because 8 appears in the position with zero elements before it.

Arrays in awk are different—they are associative. This means that each array is a collection of pairs: an index and its corresponding array element value:

     Element 3     Value 30
     Element 1     Value "foo"
     Element 0     Value 8
     Element 2     Value ""

The pairs are shown in jumbled order because their order is irrelevant.

One advantage of associative arrays is that new pairs can be added at any time. For example, suppose a tenth element is added to the array whose value is "number ten". The result is:

     Element 10    Value "number ten"
     Element 3     Value 30
     Element 1     Value "foo"
     Element 0     Value 8
     Element 2     Value ""

Now the array is sparse, which just means some indices are missing. It has elements 0–3 and 10, but doesn't have elements 4, 5, 6, 7, 8, or 9.

Another consequence of associative arrays is that the indices don't have to be positive integers. Any number, or even a string, can be an index. For example, the following is an array that translates words from English to French:

     Element "dog" Value "chien"
     Element "cat" Value "chat"
     Element "one" Value "un"
     Element 1     Value "un"

Here we decided to translate the number one in both spelled-out and numeric form—thus illustrating that a single array can have both numbers and strings as indices. In fact, array subscripts are always strings; this is discussed in more detail in Numeric Array Subscripts. Here, the number 1 isn't double-quoted, since awk automatically converts it to a string.

The value of IGNORECASE has no effect upon array subscripting. The identical string value used to store an array element must be used to retrieve it. When awk creates an array (e.g., with the split built-in function), that array's indices are consecutive integers starting at one. (See String Functions.)

awk's arrays are efficient—the time to access an element is independent of the number of elements in the array.

Next: , Previous: Array Intro, Up: Arrays

7.2 Referring to an Array Element

The principal way to use an array is to refer to one of its elements. An array reference is an expression as follows:


Here, array is the name of an array. The expression index is the index of the desired element of the array.

The value of the array reference is the current value of that array element. For example, foo[4.3] is an expression for the element of array foo at index ‘4.3’.

A reference to an array element that has no recorded value yields a value of "", the null string. This includes elements that have not been assigned any value as well as elements that have been deleted (see Delete). Such a reference automatically creates that array element, with the null string as its value. (In some cases, this is unfortunate, because it might waste memory inside awk.)

To determine whether an element exists in an array at a certain index, use the following expression:

     index in array

This expression tests whether the particular index exists, without the side effect of creating that element if it is not present. The expression has the value one (true) if array[index] exists and zero (false) if it does not exist. For example, this statement tests whether the array frequencies contains the index ‘2’:

     if (2 in frequencies)
         print "Subscript 2 is present."

Note that this is not a test of whether the array frequencies contains an element whose value is two. There is no way to do that except to scan all the elements. Also, this does not create frequencies[2], while the following (incorrect) alternative does:

     if (frequencies[2] != "")
         print "Subscript 2 is present."

Next: , Previous: Reference to Elements, Up: Arrays

7.3 Assigning Array Elements

Array elements can be assigned values just like awk variables:

     array[subscript] = value

array is the name of an array. The expression subscript is the index of the element of the array that is assigned a value. The expression value is the value to assign to that element of the array.

Next: , Previous: Assigning Elements, Up: Arrays

7.4 Basic Array Example

The following program takes a list of lines, each beginning with a line number, and prints them out in order of line number. The line numbers are not in order when they are first read—instead they are scrambled. This program sorts the lines by making an array using the line numbers as subscripts. The program then prints out the lines in sorted order of their numbers. It is a very simple program and gets confused upon encountering repeated numbers, gaps, or lines that don't begin with a number:

       if ($1 > max)
         max = $1
       arr[$1] = $0
     END {
       for (x = 1; x <= max; x++)
         print arr[x]

The first rule keeps track of the largest line number seen so far; it also stores each line into the array arr, at an index that is the line's number. The second rule runs after all the input has been read, to print out all the lines. When this program is run with the following input:

     5  I am the Five man
     2  Who are you?  The new number two!
     4  . . . And four on the floor
     1  Who is number one?
     3  I three you.

Its output is:

     1  Who is number one?
     2  Who are you?  The new number two!
     3  I three you.
     4  . . . And four on the floor
     5  I am the Five man

If a line number is repeated, the last line with a given number overrides the others. Gaps in the line numbers can be handled with an easy improvement to the program's END rule, as follows:

     END {
       for (x = 1; x <= max; x++)
         if (x in arr)
           print arr[x]

Next: , Previous: Array Example, Up: Arrays

7.5 Scanning All Elements of an Array

In programs that use arrays, it is often necessary to use a loop that executes once for each element of an array. In other languages, where arrays are contiguous and indices are limited to positive integers, this is easy: all the valid indices can be found by counting from the lowest index up to the highest. This technique won't do the job in awk, because any number or string can be an array index. So awk has a special kind of for statement for scanning an array:

     for (var in array)

This loop executes body once for each index in array that the program has previously used, with the variable var set to that index.

The following program uses this form of the for statement. The first rule scans the input records and notes which words appear (at least once) in the input, by storing a one into the array used with the word as index. The second rule scans the elements of used to find all the distinct words that appear in the input. It prints each word that is more than 10 characters long and also prints the number of such words. See String Functions, for more information on the built-in function length.

     # Record a 1 for each word that is used at least once
         for (i = 1; i <= NF; i++)
             used[$i] = 1
     # Find number of distinct words more than 10 characters long
     END {
         for (x in used)
             if (length(x) > 10) {
                 print x
         print num_long_words, "words longer than 10 characters"

See Word Sorting, for a more detailed example of this type.

The order in which elements of the array are accessed by this statement is determined by the internal arrangement of the array elements within awk and cannot be controlled or changed. This can lead to problems if new elements are added to array by statements in the loop body; it is not predictable whether the for loop will reach them. Similarly, changing var inside the loop may produce strange results. It is best to avoid such things.

Next: , Previous: Scanning an Array, Up: Arrays

7.6 The delete Statement

To remove an individual element of an array, use the delete statement:

     delete array[index]

Once an array element has been deleted, any value the element once had is no longer available. It is as if the element had never been referred to or had been given a value. The following is an example of deleting elements in an array:

     for (i in frequencies)
       delete frequencies[i]

This example removes all the elements from the array frequencies. Once an element is deleted, a subsequent for statement to scan the array does not report that element and the in operator to check for the presence of that element returns zero (i.e., false):

     delete foo[4]
     if (4 in foo)
         print "This will never be printed"

It is important to note that deleting an element is not the same as assigning it a null value (the empty string, ""). For example:

     foo[4] = ""
     if (4 in foo)
       print "This is printed, even though foo[4] is empty"

It is not an error to delete an element that does not exist. If --lint is provided on the command line (see Options), gawk issues a warning message when an element that is not in the array is deleted.

All the elements of an array may be deleted with a single statement by leaving off the subscript in the delete statement, as follows:

     delete array

This ability is a gawk extension; it is not available in compatibility mode (see Options).

Using this version of the delete statement is about three times more efficient than the equivalent loop that deletes each element one at a time.

The following statement provides a portable but nonobvious way to clear out an array:31

     split("", array)

The split function (see String Functions) clears out the target array first. This call asks it to split apart the null string. Because there is no data to split out, the function simply clears the array and then returns.

Caution: Deleting an array does not change its type; you cannot delete an array and then use the array's name as a scalar (i.e., a regular variable). For example, the following does not work:

     a[1] = 3; delete a; a = 3

Next: , Previous: Delete, Up: Arrays

7.7 Using Numbers to Subscript Arrays

An important aspect about arrays to remember is that array subscripts are always strings. When a numeric value is used as a subscript, it is converted to a string value before being used for subscripting (see Conversion). This means that the value of the built-in variable CONVFMT can affect how your program accesses elements of an array. For example:

     xyz = 12.153
     data[xyz] = 1
     CONVFMT = "%2.2f"
     if (xyz in data)
         printf "%s is in data\n", xyz
         printf "%s is not in data\n", xyz

This prints ‘12.15 is not in data’. The first statement gives xyz a numeric value. Assigning to data[xyz] subscripts data with the string value "12.153" (using the default conversion value of CONVFMT, "%.6g"). Thus, the array element data["12.153"] is assigned the value one. The program then changes the value of CONVFMT. The test ‘(xyz in data)’ generates a new string value from xyz—this time "12.15"—because the value of CONVFMT only allows two significant digits. This test fails, since "12.15" is a different string from "12.153".

According to the rules for conversions (see Conversion), integer values are always converted to strings as integers, no matter what the value of CONVFMT may happen to be. So the usual case of the following works:

     for (i = 1; i <= maxsub; i++)
         do something with array[i]

The “integer values always convert to strings as integers” rule has an additional consequence for array indexing. Octal and hexadecimal constants (see Nondecimal-numbers) are converted internally into numbers, and their original form is forgotten. This means, for example, that array[17], array[021], and array[0x11] all refer to the same element!

As with many things in awk, the majority of the time things work as one would expect them to. But it is useful to have a precise knowledge of the actual rules which sometimes can have a subtle effect on your programs.

Next: , Previous: Numeric Array Subscripts, Up: Arrays

7.8 Using Uninitialized Variables as Subscripts

Suppose it's necessary to write a program to print the input data in reverse order. A reasonable attempt to do so (with some test data) might look like this:

     $ echo 'line 1
     > line 2
     > line 3' | awk '{ l[lines] = $0; ++lines }
     > END {
     >     for (i = lines-1; i >= 0; --i)
     >        print l[i]
     > }'
     -| line 3
     -| line 2

Unfortunately, the very first line of input data did not come out in the output!

At first glance, this program should have worked. The variable lines is uninitialized, and uninitialized variables have the numeric value zero. So, awk should have printed the value of l[0].

The issue here is that subscripts for awk arrays are always strings. Uninitialized variables, when used as strings, have the value "", not zero. Thus, ‘line 1’ ends up stored in l[""]. The following version of the program works correctly:

     { l[lines++] = $0 }
     END {
         for (i = lines - 1; i >= 0; --i)
            print l[i]

Here, the ‘++’ forces lines to be numeric, thus making the “old value” numeric zero. This is then converted to "0" as the array subscript.

Even though it is somewhat unusual, the null string ("") is a valid array subscript. (d.c.) gawk warns about the use of the null string as a subscript if --lint is provided on the command line (see Options).

Next: , Previous: Uninitialized Subscripts, Up: Arrays

7.9 Multidimensional Arrays

A multidimensional array is an array in which an element is identified by a sequence of indices instead of a single index. For example, a two-dimensional array requires two indices. The usual way (in most languages, including awk) to refer to an element of a two-dimensional array named grid is with grid[x,y].

Multidimensional arrays are supported in awk through concatenation of indices into one string. awk converts the indices into strings (see Conversion) and concatenates them together, with a separator between them. This creates a single string that describes the values of the separate indices. The combined string is used as a single index into an ordinary, one-dimensional array. The separator used is the value of the built-in variable SUBSEP.

For example, suppose we evaluate the expression ‘foo[5,12] = "value"’ when the value of SUBSEP is "@". The numbers 5 and 12 are converted to strings and concatenated with an ‘@’ between them, yielding "5@12"; thus, the array element foo["5@12"] is set to "value".

Once the element's value is stored, awk has no record of whether it was stored with a single index or a sequence of indices. The two expressions ‘foo[5,12]’ and ‘foo[5 SUBSEP 12] are always equivalent.

The default value of SUBSEP is the string "\034", which contains a nonprinting character that is unlikely to appear in an awk program or in most input data. The usefulness of choosing an unlikely character comes from the fact that index values that contain a string matching SUBSEP can lead to combined strings that are ambiguous. Suppose that SUBSEP is "@"; then ‘foo["a@b", "c"] and ‘foo["a", "b@c"] are indistinguishable because both are actually stored as ‘foo["a@b@c"]’.

To test whether a particular index sequence exists in a multidimensional array, use the same operator (‘in’) that is used for single dimensional arrays. Write the whole sequence of indices in parentheses, separated by commas, as the left operand:

     (subscript1, subscript2, ...) in array

The following example treats its input as a two-dimensional array of fields; it rotates this array 90 degrees clockwise and prints the result. It assumes that all lines have the same number of elements:

          if (max_nf < NF)
               max_nf = NF
          max_nr = NR
          for (x = 1; x <= NF; x++)
               vector[x, NR] = $x
     END {
          for (x = 1; x <= max_nf; x++) {
               for (y = max_nr; y >= 1; --y)
                    printf("%s ", vector[x, y])

When given the input:

     1 2 3 4 5 6
     2 3 4 5 6 1
     3 4 5 6 1 2
     4 5 6 1 2 3

the program produces the following output:

     4 3 2 1
     5 4 3 2
     6 5 4 3
     1 6 5 4
     2 1 6 5
     3 2 1 6

Next: , Previous: Multi-dimensional, Up: Arrays

7.10 Scanning Multidimensional Arrays

There is no special for statement for scanning a “multidimensional” array. There cannot be one, because, in truth, there are no multidimensional arrays or elements—there is only a multidimensional way of accessing an array.

However, if your program has an array that is always accessed as multidimensional, you can get the effect of scanning it by combining the scanning for statement (see Scanning an Array) with the built-in split function (see String Functions). It works in the following manner:

     for (combined in array) {
         split(combined, separate, SUBSEP)

This sets the variable combined to each concatenated combined index in the array, and splits it into the individual indices by breaking it apart where the value of SUBSEP appears. The individual indices then become the elements of the array separate.

Thus, if a value is previously stored in array[1, "foo"]; then an element with index "1\034foo" exists in array. (Recall that the default value of SUBSEP is the character with code 034.) Sooner or later, the for statement finds that index and does an iteration with the variable combined set to "1\034foo". Then the split function is called as follows:

     split("1\034foo", separate, "\034")

The result is to set separate[1] to "1" and separate[2] to "foo". Presto! The original sequence of separate indices is recovered.

Previous: Multi-scanning, Up: Arrays

7.11 Sorting Array Values and Indices with gawk

The order in which an array is scanned with a ‘for (i in array)’ loop is essentially arbitrary. In most awk implementations, sorting an array requires writing a sort function. While this can be educational for exploring different sorting algorithms, usually that's not the point of the program. gawk provides the built-in asort and asorti functions (see String Functions) for sorting arrays. For example:

     populate the array data
     n = asort(data)
     for (i = 1; i <= n; i++)
         do something with data[i]

After the call to asort, the array data is indexed from 1 to some number n, the total number of elements in data. (This count is asort's return value.) data[1] <= data[2] <= data[3], and so on. The comparison of array elements is done using gawk's usual comparison rules (see Typing and Comparison).

An important side effect of calling asort is that the array's original indices are irrevocably lost. As this isn't always desirable, asort accepts a second argument:

     populate the array source
     n = asort(source, dest)
     for (i = 1; i <= n; i++)
         do something with dest[i]

In this case, gawk copies the source array into the dest array and then sorts dest, destroying its indices. However, the source array is not affected.

Often, what's needed is to sort on the values of the indices instead of the values of the elements. To do that, starting with gawk 3.1.2, use the asorti function. The interface is identical to that of asort, except that the index values are used for sorting, and become the values of the result array:

     { source[$0] = some_func($0) }
     END {
         n = asorti(source, dest)
         for (i = 1; i <= n; i++) {
             do something with dest[i]             Work with sorted indices directly
             do something with source[dest[i]]     Access original array via sorted indices

If your version of gawk is 3.1.0 or 3.1.1, you don't have asorti. Instead, use a helper array to hold the sorted index values, and then access the original array's elements. It works in the following way:

     populate the array data
     # copy indices
     j = 1
     for (i in data) {
         ind[j] = i    # index value becomes element value
     n = asort(ind)    # index values are now sorted
     for (i = 1; i <= n; i++) {
         do something with ind[i]           Work with sorted indices directly
         do something with data[ind[i]]     Access original array via sorted indices

Sorting the array by replacing the indices provides maximal flexibility. To traverse the elements in decreasing order, use a loop that goes from n down to 1, either over the elements or over the indices.

Copying array indices and elements isn't expensive in terms of memory. Internally, gawk maintains reference counts to data. For example, when asort copies the first array to the second one, there is only one copy of the original array elements' data, even though both arrays use the values. Similarly, when copying the indices from data to ind, there is only one copy of the actual index strings.

We said previously that comparisons are done using gawk's “usual comparison rules.” Because IGNORECASE affects string comparisons, the value of IGNORECASE also affects sorting for both asort and asorti. Caveat Emptor.

Next: , Previous: Arrays, Up: Top

8 Functions

This chapter describes awk's built-in functions, which fall into three categories: numeric, string, and I/O. gawk provides additional groups of functions to work with values that represent time, do bit manipulation, and internationalize and localize programs.

Besides the built-in functions, awk has provisions for writing new functions that the rest of a program can use. The second half of this chapter describes these user-defined functions.

Next: , Up: Functions

8.1 Built-in Functions

Built-in functions are always available for your awk program to call. This section defines all the built-in functions in awk; some of these are mentioned in other sections but are summarized here for your convenience.

Next: , Up: Built-in

8.1.1 Calling Built-in Functions

To call one of awk's built-in functions, write the name of the function followed by arguments in parentheses. For example, ‘atan2(y + z, 1)’ is a call to the function atan2 and has two arguments.

Whitespace is ignored between the built-in function name and the open parenthesis, and it is good practice to avoid using whitespace there. User-defined functions do not permit whitespace in this way, and it is easier to avoid mistakes by following a simple convention that always works—no whitespace after a function name.

Each built-in function accepts a certain number of arguments. In some cases, arguments can be omitted. The defaults for omitted arguments vary from function to function and are described under the individual functions. In some awk implementations, extra arguments given to built-in functions are ignored. However, in gawk, it is a fatal error to give extra arguments to a built-in function.

When a function is called, expressions that create the function's actual parameters are evaluated completely before the call is performed. For example, in the following code fragment:

     i = 4
     j = sqrt(i++)

the variable i is incremented to the value five before sqrt is called with a value of four for its actual parameter. The order of evaluation of the expressions used for the function's parameters is undefined. Thus, avoid writing programs that assume that parameters are evaluated from left to right or from right to left. For example:

     i = 5
     j = atan2(i++, i *= 2)

If the order of evaluation is left to right, then i first becomes 6, and then 12, and atan2 is called with the two arguments 6 and 12. But if the order of evaluation is right to left, i first becomes 10, then 11, and atan2 is called with the two arguments 11 and 10.

Next: , Previous: Calling Built-in, Up: Built-in

8.1.2 Numeric Functions

The following list describes all of the built-in functions that work with numbers. Optional parameters are enclosed in square brackets ([ ]):

This returns the nearest integer to x, located between x and zero and truncated toward zero.

For example, int(3) is 3, int(3.9) is 3, int(-3.9) is −3, and int(-3) is −3 as well.

This returns the positive square root of x. gawk reports an error if x is negative. Thus, sqrt(4) is 2.
This returns the exponential of x (e ^ x) or reports an error if x is out of range. The range of values x can have depends on your machine's floating-point representation.
This returns the natural logarithm of x, if x is positive; otherwise, it reports an error.
This returns the sine of x, with x in radians.
This returns the cosine of x, with x in radians.
atan2(y, x)
This returns the arctangent of y / x in radians.
This returns a random number. The values of rand are uniformly distributed between zero and one. The value could be zero but is never one.32

Often random integers are needed instead. Following is a user-defined function that can be used to obtain a random non-negative integer less than n:

          function randint(n) {
               return int(n * rand())

The multiplication produces a random number greater than zero and less than n. Using int, this result is made into an integer between zero and n − 1, inclusive.

The following example uses a similar function to produce random integers between one and n. This program prints a new random number for each input record:

          # Function to roll a simulated die.
          function roll(n) { return 1 + int(rand() * n) }
          # Roll 3 six-sided dice and
          # print total number of points.
                printf("%d points\n",

Caution: In most awk implementations, including gawk, rand starts generating numbers from the same starting number, or seed, each time you run awk. Thus, a program generates the same results each time you run it. The numbers are random within one awk run but predictable from run to run. This is convenient for debugging, but if you want a program to do different things each time it is used, you must change the seed to a value that is different in each run. To do this, use srand.

The function srand sets the starting point, or seed, for generating random numbers to the value x.

Each seed value leads to a particular sequence of random numbers.33 Thus, if the seed is set to the same value a second time, the same sequence of random numbers is produced again.

Different awk implementations use different random-number generators internally. Don't expect the same awk program to produce the same series of random numbers when executed by different versions of awk.

If the argument x is omitted, as in ‘srand()’, then the current date and time of day are used for a seed. This is the way to get random numbers that are truly unpredictable.

The return value of srand is the previous seed. This makes it easy to keep track of the seeds in case you need to consistently reproduce sequences of random numbers.

Next: , Previous: Numeric Functions, Up: Built-in

8.1.3 String-Manipulation Functions

The functions in this section look at or change the text of one or more strings. Optional parameters are enclosed in square brackets ([ ]). Those functions that are specific to gawk are marked with a pound sign (‘#’):

asort(source [, dest]) #
asort is a gawk-specific extension, returning the number of elements in the array source. The contents of source are sorted using gawk's normal rules for comparing values (in particular, IGNORECASE affects the sorting) and the indices of the sorted values of source are replaced with sequential integers starting with one. If the optional array dest is specified, then source is duplicated into dest. dest is then sorted, leaving the indices of source unchanged. For example, if the contents of a are as follows:
          a["last"] = "de"
          a["first"] = "sac"
          a["middle"] = "cul"

A call to asort:


results in the following contents of a:

          a[1] = "cul"
          a[2] = "de"
          a[3] = "sac"

The asort function is described in more detail in Array Sorting. asort is a gawk extension; it is not available in compatibility mode (see Options).

asorti(source [, dest]) #
asorti is a gawk-specific extension, returning the number of elements in the array source. It works similarly to asort, however, the indices are sorted, instead of the values. As array indices are always strings, the comparison performed is always a string comparison. (Here too, IGNORECASE affects the sorting.)

The asorti function is described in more detail in Array Sorting. It was added in gawk 3.1.2. asorti is a gawk extension; it is not available in compatibility mode (see Options).

index(in, find)
This searches the string in for the first occurrence of the string find, and returns the position in characters where that occurrence begins in the string in. Consider the following example:
          $ awk 'BEGIN { print index("peanut", "an") }'
          -| 3

If find is not found, index returns zero. (Remember that string indices in awk start at one.)

This returns the number of characters in string. If string is a number, the length of the digit string representing that number is returned. For example, length("abcde") is 5. By contrast, length(15 * 35) works out to 3. In this example, 15 * 35 = 525, and 525 is then converted to the string "525", which has three characters.

If no argument is supplied, length returns the length of $0.

NOTE: In older versions of awk, the length function could be called without any parentheses. Doing so is marked as “deprecated” in the POSIX standard. This means that while a program can do this, it is a feature that can eventually be removed from a future version of the standard. Therefore, for programs to be maximally portable, always supply the parentheses.

Beginning with gawk version 3.2, when supplied an array argument, the length function returns the number of elements in the array. This is less useful than it might seem at first, as the array is not guaranteed to be indexed from one to the number of elements in it. If --lint is provided on the command line (see Options), gawk warns that passing an array argument is not portable. If --posix is supplied, using an array argument is a fatal error (see Arrays).

match(string, regexp [, array])
The match function searches string for the longest, leftmost substring matched by the regular expression, regexp. It returns the character position, or index, at which that substring begins (one, if it starts at the beginning of string). If no match is found, it returns zero.

The regexp argument may be either a regexp constant (‘/.../’) or a string constant ("..."). In the latter case, the string is treated as a regexp to be matched. Computed Regexps, for a discussion of the difference between the two forms, and the implications for writing your program correctly.

The order of the first two arguments is backwards from most other string functions that work with regular expressions, such as sub and gsub. It might help to remember that for match, the order is the same as for the ‘~’ operator: ‘string ~ regexp’.

The match function sets the built-in variable RSTART to the index. It also sets the built-in variable RLENGTH to the length in characters of the matched substring. If no match is found, RSTART is set to zero, and RLENGTH to −1.

For example:

                 if ($1 == "FIND")
                   regex = $2
                 else {
                   where = match($0, regex)
                   if (where != 0)
                     print "Match of", regex, "found at",
                               where, "in", $0

This program looks for lines that match the regular expression stored in the variable regex. This regular expression can be changed. If the first word on a line is ‘FIND’, regex is changed to be the second word on that line. Therefore, if given:

          FIND ru+n
          My program runs
          but not very quickly
          FIND Melvin
          This line is property of Reality Engineering Co.
          Melvin was here.

awk prints:

          Match of ru+n found at 12 in My program runs
          Match of Melvin found at 1 in Melvin was here.

If array is present, it is cleared, and then the 0th element of array is set to the entire portion of string matched by regexp. If regexp contains parentheses, the integer-indexed elements of array are set to contain the portion of string matching the corresponding parenthesized subexpression. For example:

          $ echo foooobazbarrrrr |
          > gawk '{ match($0, /(fo+).+(bar*)/, arr)
          >           print arr[1], arr[2] }'
          -| foooo barrrrr

In addition, beginning with gawk 3.1.2, multidimensional subscripts are available providing the start index and length of each matched subexpression:

          $ echo foooobazbarrrrr |
          > gawk '{ match($0, /(fo+).+(bar*)/, arr)
          >           print arr[1], arr[2]
          >           print arr[1, "start"], arr[1, "length"]
          >           print arr[2, "start"], arr[2, "length"]
          > }'
          -| foooo barrrrr
          -| 1 5
          -| 9 7

There may not be subscripts for the start and index for every parenthesized subexpressions, since they may not all have matched text; thus they should be tested for with the in operator (see Reference to Elements).

The array argument to match is a gawk extension. In compatibility mode (see Options), using a third argument is a fatal error.

split(string, array [, fieldsep])
This function divides string into pieces separated by fieldsep and stores the pieces in array. The first piece is stored in array[1], the second piece in array[2], and so forth. The string value of the third argument, fieldsep, is a regexp describing where to split string (much as FS can be a regexp describing where to split input records). If fieldsep is omitted, the value of FS is used. split returns the number of elements created.

The split function splits strings into pieces in a manner similar to the way input lines are split into fields. For example:

          split("cul-de-sac", a, "-")

splits the string ‘cul-de-sac’ into three fields using ‘-’ as the separator. It sets the contents of the array a as follows:

          a[1] = "cul"
          a[2] = "de"
          a[3] = "sac"

The value returned by this call to split is three.

As with input field-splitting, when the value of fieldsep is " ", leading and trailing whitespace is ignored, and the elements are separated by runs of whitespace. Also as with input field-splitting, if fieldsep is the null string, each individual character in the string is split into its own array element. (This is a gawk-specific extension.)

Note, however, that RS has no effect on the way split works. Even though ‘RS = ""’ causes newline to also be an input field separator, this does not affect how split splits strings.

Modern implementations of awk, including gawk, allow the third argument to be a regexp constant (/abc/) as well as a string. (d.c.) The POSIX standard allows this as well. Computed Regexps, for a discussion of the difference between using a string constant or a regexp constant, and the implications for writing your program correctly.

Before splitting the string, split deletes any previously existing elements in the array array.

If string is null, the array has no elements. (So this is a portable way to delete an entire array with one statement. See Delete.)

If string does not match fieldsep at all (but is not null), array has one element only. The value of that element is the original string.

sprintf(format, expression1, ...)
This returns (without printing) the string that printf would have printed out with the same arguments (see Printf). For example:
          pival = sprintf("pi = %.2f (approx.)", 22/7)

assigns the string "pi = 3.14 (approx.)" to the variable pival.

strtonum(str) #
Examines str and returns its numeric value. If str begins with a leading ‘0’, strtonum assumes that str is an octal number. If str begins with a leading ‘0x’ or ‘0X’, strtonum assumes that str is a hexadecimal number. For example:
          $ echo 0x11 |
          > gawk '{ printf "%d\n", strtonum($1) }'
          -| 17

Using the strtonum function is not the same as adding zero to a string value; the automatic coercion of strings to numbers works only for decimal data, not for octal or hexadecimal.34

Note also that strtonum uses the current locale's decimal point for recognizing numbers.

strtonum is a gawk extension; it is not available in compatibility mode (see Options).

sub(regexp, replacement [, target])
The sub function alters the value of target. It searches this value, which is treated as a string, for the leftmost, longest substring matched by the regular expression regexp. Then the entire string is changed by replacing the matched text with replacement. The modified string becomes the new value of target.

The regexp argument may be either a regexp constant (‘/.../’) or a string constant ("..."). In the latter case, the string is treated as a regexp to be matched. Computed Regexps, for a discussion of the difference between the two forms, and the implications for writing your program correctly.

This function is peculiar because target is not simply used to compute a value, and not just any expression will do—it must be a variable, field, or array element so that sub can store a modified value there. If this argument is omitted, then the default is to use and alter $0.35 For example:

          str = "water, water, everywhere"
          sub(/at/, "ith", str)

sets str to "wither, water, everywhere", by replacing the leftmost longest occurrence of ‘at’ with ‘ith’.

The sub function returns the number of substitutions made (either one or zero).

If the special character ‘&’ appears in replacement, it stands for the precise substring that was matched by regexp. (If the regexp can match more than one string, then this precise substring may vary.) For example:

          { sub(/candidate/, "& and his wife"); print }

changes the first occurrence of ‘candidate’ to ‘candidate and his wife’ on each input line. Here is another example:

          $ awk 'BEGIN {
          >         str = "daabaaa"
          >         sub(/a+/, "C&C", str)
          >         print str
          > }'
          -| dCaaCbaaa

This shows how ‘&’ can represent a nonconstant string and also illustrates the “leftmost, longest” rule in regexp matching (see Leftmost Longest).

The effect of this special character (‘&’) can be turned off by putting a backslash before it in the string. As usual, to insert one backslash in the string, you must write two backslashes. Therefore, write ‘\\&’ in a string constant to include a literal ‘&’ in the replacement. For example, the following shows how to replace the first ‘|’ on each line with an ‘&’:

          { sub(/\|/, "\\&"); print }

As mentioned, the third argument to sub must be a variable, field or array reference. Some versions of awk allow the third argument to be an expression that is not an lvalue. In such a case, sub still searches for the pattern and returns zero or one, but the result of the substitution (if any) is thrown away because there is no place to put it. Such versions of awk accept expressions such as the following:

          sub(/USA/, "United States", "the USA and Canada")

For historical compatibility, gawk accepts erroneous code, such as in the previous example. However, using any other nonchangeable object as the third parameter causes a fatal error and your program will not run.

Finally, if the regexp is not a regexp constant, it is converted into a string, and then the value of that string is treated as the regexp to match.

gsub(regexp, replacement [, target])
This is similar to the sub function, except gsub replaces all of the longest, leftmost, nonoverlapping matching substrings it can find. The ‘g’ in gsub stands for “global,” which means replace everywhere. For example:
          { gsub(/Britain/, "United Kingdom"); print }

replaces all occurrences of the string ‘Britain’ with ‘United Kingdom’ for all input records.

The gsub function returns the number of substitutions made. If the variable to search and alter (target) is omitted, then the entire input record ($0) is used. As in sub, the characters ‘&’ and ‘\’ are special, and the third argument must be assignable.

gensub(regexp, replacement, how [, target]) #
gensub is a general substitution function. Like sub and gsub, it searches the target string target for matches of the regular expression regexp. Unlike sub and gsub, the modified string is returned as the result of the function and the original target string is not changed. If how is a string beginning with ‘g’ or ‘G’, then it replaces all matches of regexp with replacement. Otherwise, how is treated as a number that indicates which match of regexp to replace. If no target is supplied, $0 is used.

gensub provides an additional feature that is not available in sub or gsub: the ability to specify components of a regexp in the replacement text. This is done by using parentheses in the regexp to mark the components and then specifying ‘\N’ in the replacement text, where N is a digit from 1 to 9. For example:

          $ gawk '
          > BEGIN {
          >      a = "abc def"
          >      b = gensub(/(.+) (.+)/, "\\2 \\1", "g", a)
          >      print b
          > }'
          -| def abc

As with sub, you must type two backslashes in order to get one into the string. In the replacement text, the sequence ‘\0’ represents the entire matched text, as does the character ‘&’.

The following example shows how you can use the third argument to control which match of the regexp should be changed:

          $ echo a b c a b c |
          > gawk '{ print gensub(/a/, "AA", 2) }'
          -| a b c AA b c

In this case, $0 is used as the default target string. gensub returns the new string as its result, which is passed directly to print for printing.

If the how argument is a string that does not begin with ‘g’ or ‘G’, or if it is a number that is less than or equal to zero, only one substitution is performed. If how is zero, gawk issues a warning message.

If regexp does not match target, gensub's return value is the original unchanged value of target.

gensub is a gawk extension; it is not available in compatibility mode (see Options).

substr(string, start [, length])
This returns a length-character-long substring of string, starting at character number start. The first character of a string is character number one.36 For example, substr("washington", 5, 3) returns "ing".

If length is not present, this function returns the whole suffix of string that begins at character number start. For example, substr("washington", 5) returns "ington". The whole suffix is also returned if length is greater than the number of characters remaining in the string, counting from character start.

If start is less than one, substr treats it as if it was one. (POSIX doesn't specify what to do in this case: Unix awk acts this way, and therefore gawk does too.) If start is greater than the number of characters in the string, substr returns the null string. Similarly, if length is present but less than or equal to zero, the null string is returned.

The string returned by substr cannot be assigned. Thus, it is a mistake to attempt to change a portion of a string, as shown in the following example:

          string = "abcdef"
          # try to get "abCDEf", won't work
          substr(string, 3, 3) = "CDE"

It is also a mistake to use substr as the third argument of sub or gsub:

          gsub(/xyz/, "pdq", substr($0, 5, 20))  # WRONG

(Some commercial versions of awk do in fact let you use substr this way, but doing so is not portable.)

If you need to replace bits and pieces of a string, combine substr with string concatenation, in the following manner:

          string = "abcdef"
          string = substr(string, 1, 2) "CDE" substr(string, 6)

This returns a copy of string, with each uppercase character in the string replaced with its corresponding lowercase character. Nonalphabetic characters are left unchanged. For example, tolower("MiXeD cAsE 123") returns "mixed case 123".
This returns a copy of string, with each lowercase character in the string replaced with its corresponding uppercase character. Nonalphabetic characters are left unchanged. For example, toupper("MiXeD cAsE 123") returns "MIXED CASE 123".

Up: String Functions More About ‘\’ and ‘&’ with sub, gsub, and gensub

When using sub, gsub, or gensub, and trying to get literal backslashes and ampersands into the replacement text, you need to remember that there are several levels of escape processing going on.

First, there is the lexical level, which is when awk reads your program and builds an internal copy of it that can be executed. Then there is the runtime level, which is when awk actually scans the replacement string to determine what to generate.

At both levels, awk looks for a defined set of characters that can come after a backslash. At the lexical level, it looks for the escape sequences listed in Escape Sequences. Thus, for every ‘\’ that awk processes at the runtime level, type two backslashes at the lexical level. When a character that is not valid for an escape sequence follows the ‘\’, Unix awk and gawk both simply remove the initial ‘\’ and put the next character into the string. Thus, for example, "a\qb" is treated as "aqb".

At the runtime level, the various functions handle sequences of ‘\’ and ‘&’ differently. The situation is (sadly) somewhat complex. Historically, the sub and gsub functions treated the two character sequence ‘\&’ specially; this sequence was replaced in the generated text with a single ‘&’. Any other ‘\’ within the replacement string that did not precede an ‘&’ was passed through unchanged. This is illustrated in table-sub-escapes.

      You type         sub sees          sub generates
      ———–         ————–          ———————
          \&              &            the matched text
         \\&             \&            a literal ‘&\\\&             \&            a literal ‘&\\\\&            \\&            a literal ‘\&\\\\\&            \\&            a literal ‘\&\\\\\\&           \\\&            a literal ‘\\&\\q             \q            a literal ‘\q

Table 8.1: Historical Escape Sequence Processing for sub and gsub

This table shows both the lexical-level processing, where an odd number of backslashes becomes an even number at the runtime level, as well as the runtime processing done by sub. (For the sake of simplicity, the rest of the following tables only show the case of even numbers of backslashes entered at the lexical level.)

The problem with the historical approach is that there is no way to get a literal ‘\’ followed by the matched text.

The 1992 POSIX standard attempted to fix this problem. That standard says that sub and gsub look for either a ‘\’ or an ‘&’ after the ‘\’. If either one follows a ‘\’, that character is output literally. The interpretation of ‘\’ and ‘&’ then becomes as shown in table-sub-posix-92.

      You type         sub sees          sub generates
      ———–         ————–          ———————
           &              &            the matched text
         \\&             \&            a literal ‘&\\\\&            \\&            a literal ‘\’, then the matched text
     \\\\\\&           \\\&            a literal ‘\&

Table 8.2: 1992 POSIX Rules for sub and gsub Escape Sequence Processing

This appears to solve the problem. Unfortunately, the phrasing of the standard is unusual. It says, in effect, that ‘\’ turns off the special meaning of any following character, but for anything other than ‘\’ and ‘&’, such special meaning is undefined. This wording leads to two problems:

Because of the problems just listed, in 1996, the gawk maintainer submitted proposed text for a revised standard that reverts to rules that correspond more closely to the original existing practice. The proposed rules have special cases that make it possible to produce a ‘\’ preceding the matched text. This is shown in table-sub-proposed.

      You type         sub sees         sub generates
      ———–         ————–         ———————
     \\\\\\&           \\\&            a literal ‘\&\\\\&            \\&            a literal ‘\’, followed by the matched text
         \\&             \&            a literal ‘&\\q             \q            a literal ‘\q\\\\             \\            \\

Table 8.3: Proposed rules for sub and backslash

In a nutshell, at the runtime level, there are now three special sequences of characters (‘\\\&’, ‘\\&’ and ‘\&’) whereas historically there was only one. However, as in the historical case, any ‘\’ that is not part of one of these three sequences is not special and appears in the output literally.

gawk 3.0 and 3.1 follow these proposed POSIX rules for sub and gsub. The POSIX standard took much longer to be revised than was expected in 1996. The 2001 standard does not follow the above rules. Instead, the rules there are somewhat simpler. The results are similar except for one case.

The 2001 POSIX rules state that ‘\&’ in the replacement string produces a literal ‘&’, ‘\\’ produces a literal ‘\’, and ‘\’ followed by anything else is not special; the ‘\’ is placed straight into the output. These rules are presented in table-posix-2001-sub.

      You type         sub sees         sub generates
      ———–         ————–         ———————
     \\\\\\&           \\\&            a literal ‘\&\\\\&            \\&            a literal ‘\’, followed by the matched text
         \\&             \&            a literal ‘&\\q             \q            a literal ‘\q\\\\             \\            \

Table 8.4: POSIX 2001 rules for sub

The only case where the difference is noticeable is the last one: ‘\\\\’ is seen as ‘\\’ and produces ‘\’ instead of ‘\\’.

Starting with version 3.1.4, gawk follows the POSIX rules when --posix is specified (see Options). Otherwise, it continues to follow the 1996 proposed rules, since, as of this writing, that has been its behavior for over seven years.

NOTE: At the next major release, gawk will switch to using the POSIX 2001 rules by default.

The rules for gensub are considerably simpler. At the runtime level, whenever gawk sees a ‘\’, if the following character is a digit, then the text that matched the corresponding parenthesized subexpression is placed in the generated output. Otherwise, no matter what character follows the ‘\’, it appears in the generated text and the ‘\’ does not, as shown in table-gensub-escapes.

       You type          gensub sees         gensub generates
       ———–          ——————         ————————–
           &                    &            the matched text
         \\&                   \&            a literal ‘&\\\\                   \\            a literal ‘\\\\\&                  \\&            a literal ‘\’, then the matched text
     \\\\\\&                 \\\&            a literal ‘\&\\q                   \q            a literal ‘q

Table 8.5: Escape Sequence Processing for gensub

Because of the complexity of the lexical and runtime level processing and the special cases for sub and gsub, we recommend the use of gawk and gensub when you have to do substitutions.

Advanced Notes: Matching the Null String

In awk, the ‘*’ operator can match the null string. This is particularly important for the sub, gsub, and gensub functions. For example:

     $ echo abc | awk '{ gsub(/m*/, "X"); print }'
     -| XaXbXcX

Although this makes a certain amount of sense, it can be surprising.

Next: , Previous: String Functions, Up: Built-in

8.1.4 Input/Output Functions

The following functions relate to input/output (I/O). Optional parameters are enclosed in square brackets ([ ]):

close(filename [, how])
Close the file filename for input or output. Alternatively, the argument may be a shell command that was used for creating a coprocess, or for redirecting to or from a pipe; then the coprocess or pipe is closed. See Close Files And Pipes, for more information.

When closing a coprocess, it is occasionally useful to first close one end of the two-way pipe and then to close the other. This is done by providing a second argument to close. This second argument should be one of the two string values "to" or "from", indicating which end of the pipe to close. Case in the string does not matter. See Two-way I/O, which discusses this feature in more detail and gives an example.

Flush any buffered output associated with filename, which is either a file opened for writing or a shell command for redirecting output to a pipe or coprocess.

Many utility programs buffer their output; i.e., they save information to write to a disk file or terminal in memory until there is enough for it to be worthwhile to send the data to the output device. This is often more efficient than writing every little bit of information as soon as it is ready. However, sometimes it is necessary to force a program to flush its buffers; that is, write the information to its destination, even if a buffer is not full. This is the purpose of the fflush function—gawk also buffers its output and the fflush function forces gawk to flush its buffers.

fflush was added to the Bell Laboratories research version of awk in 1994; it is not part of the POSIX standard and is not available if --posix has been specified on the command line (see Options).

gawk extends the fflush function in two ways. The first is to allow no argument at all. In this case, the buffer for the standard output is flushed. The second is to allow the null string ("") as the argument. In this case, the buffers for all open output files and pipes are flushed.

fflush returns zero if the buffer is successfully flushed; otherwise, it returns −1. In the case where all buffers are flushed, the return value is zero only if all buffers were flushed successfully. Otherwise, it is −1, and gawk warns about the problem filename.

gawk also issues a warning message if you attempt to flush a file or pipe that was opened for reading (such as with getline), or if filename is not an open file, pipe, or coprocess. In such a case, fflush returns −1, as well.

Executes operating-system commands and then returns to the awk program. The system function executes the command given by the string command. It returns the status returned by the command that was executed as its value.

For example, if the following fragment of code is put in your awk program:

          END {
               system("date | mail -s 'awk run done' root")

the system administrator is sent mail when the awk program finishes processing input and begins its end-of-input processing.

Note that redirecting print or printf into a pipe is often enough to accomplish your task. If you need to run many commands, it is more efficient to simply print them down a pipeline to the shell:

          while (more stuff to do)
              print command | "/bin/sh"

However, if your awk program is interactive, system is useful for cranking up large self-contained programs, such as a shell or an editor. Some operating systems cannot implement the system function. system causes a fatal error if it is not supported.

Advanced Notes: Interactive Versus Noninteractive Buffering

As a side point, buffering issues can be even more confusing, depending upon whether your program is interactive, i.e., communicating with a user sitting at a keyboard.38

Interactive programs generally line buffer their output; i.e., they write out every line. Noninteractive programs wait until they have a full buffer, which may be many lines of output. Here is an example of the difference:

     $ awk '{ print $1 + $2 }'
     1 1
     -| 2
     2 3
     -| 5

Each line of output is printed immediately. Compare that behavior with this example:

     $ awk '{ print $1 + $2 }' | cat
     1 1
     2 3
     -| 2
     -| 5

Here, no output is printed until after the Ctrl-d is typed, because it is all buffered and sent down the pipe to cat in one shot.

Advanced Notes: Controlling Output Buffering with system

The fflush function provides explicit control over output buffering for individual files and pipes. However, its use is not portable to many other awk implementations. An alternative method to flush output buffers is to call system with a null string as its argument:

     system("")   # flush output

gawk treats this use of the system function as a special case and is smart enough not to run a shell (or other command interpreter) with the empty command. Therefore, with gawk, this idiom is not only useful, it is also efficient. While this method should work with other awk implementations, it does not necessarily avoid starting an unnecessary shell. (Other implementations may only flush the buffer associated with the standard output and not necessarily all buffered output.)

If you think about what a programmer expects, it makes sense that system should flush any pending output. The following program:

     BEGIN {
          print "first print"
          system("echo system echo")
          print "second print"

must print:

     first print
     system echo
     second print

and not:

     system echo
     first print
     second print

If awk did not flush its buffers before calling system, you would see the latter (undesirable) output.

Next: , Previous: I/O Functions, Up: Built-in

8.1.5 Using gawk's Timestamp Functions

awk programs are commonly used to process log files containing timestamp information, indicating when a particular log record was written. Many programs log their timestamp in the form returned by the time system call, which is the number of seconds since a particular epoch. On POSIX-compliant systems, it is the number of seconds since 1970-01-01 00:00:00 UTC, not counting leap seconds.39 All known POSIX-compliant systems support timestamps from 0 through 2^31 - 1, which is sufficient to represent times through 2038-01-19 03:14:07 UTC. Many systems support a wider range of timestamps, including negative timestamps that represent times before the epoch.

In order to make it easier to process such log files and to produce useful reports, gawk provides the following functions for working with timestamps. They are gawk extensions; they are not specified in the POSIX standard, nor are they in any other known version of awk.40 Optional parameters are enclosed in square brackets ([ ]):

This function returns the current time as the number of seconds since the system epoch. On POSIX systems, this is the number of seconds since 1970-01-01 00:00:00 UTC, not counting leap seconds. It may be a different number on other systems.
This function turns datespec into a timestamp in the same form as is returned by systime. It is similar to the function of the same name in ISO C. The argument, datespec, is a string of the form "YYYY MM DD HH MM SS [DST]". The string consists of six or seven numbers representing, respectively, the full year including century, the month from 1 to 12, the day of the month from 1 to 31, the hour of the day from 0 to 23, the minute from 0 to 59, the second from 0 to 60,41 and an optional daylight-savings flag.

The values of these numbers need not be within the ranges specified; for example, an hour of −1 means 1 hour before midnight. The origin-zero Gregorian calendar is assumed, with year 0 preceding year 1 and year −1 preceding year 0. The time is assumed to be in the local timezone. If the daylight-savings flag is positive, the time is assumed to be daylight savings time; if zero, the time is assumed to be standard time; and if negative (the default), mktime attempts to determine whether daylight savings time is in effect for the specified time.

If datespec does not contain enough elements or if the resulting time is out of range, mktime returns −1.

strftime([format [, timestamp [, utc-flag]]])
This function returns a string. It is similar to the function of the same name in ISO C. The time specified by timestamp is used to produce a string, based on the contents of the format string. If utc-flag is present and is either non-zero or non-null, the value is formatted as UTC (Coordinated Universal Time, formerly GMT or Greenwich Mean Time). Otherwise, the value is formatted for the local time zone. The timestamp is in the same format as the value returned by the systime function. If no timestamp argument is supplied, gawk uses the current time of day as the timestamp. If no format argument is supplied, strftime uses "%a %b %d %H:%M:%S %Z %Y". This format string produces output that is (almost) equivalent to that of the date utility. (Versions of gawk prior to 3.0 require the format argument.)

The systime function allows you to compare a timestamp from a log file with the current time of day. In particular, it is easy to determine how long ago a particular record was logged. It also allows you to produce log records using the “seconds since the epoch” format.

The mktime function allows you to convert a textual representation of a date and time into a timestamp. This makes it easy to do before/after comparisons of dates and times, particularly when dealing with date and time data coming from an external source, such as a log file.

The strftime function allows you to easily turn a timestamp into human-readable information. It is similar in nature to the sprintf function (see String Functions), in that it copies nonformat specification characters verbatim to the returned string, while substituting date and time values for format specifications in the format string.

strftime is guaranteed by the 1999 ISO C standard42 to support the following date format specifications:

The locale's abbreviated weekday name.
The locale's full weekday name.
The locale's abbreviated month name.
The locale's full month name.
The locale's “appropriate” date and time representation. (This is ‘%A %B %d %T %Y’ in the "C" locale.)
The century. This is the year divided by 100 and truncated to the next lower integer.
The day of the month as a decimal number (01–31).
Equivalent to specifying ‘%m/%d/%y’.
The day of the month, padded with a space if it is only one digit.
Equivalent to specifying ‘%Y-%m-%d’. This is the ISO 8601 date format.
The year modulo 100 of the ISO week number, as a decimal number (00–99). For example, January 1, 1993 is in week 53 of 1992. Thus, the year of its ISO week number is 1992, even though its year is 1993. Similarly, December 31, 1973 is in week 1 of 1974. Thus, the year of its ISO week number is 1974, even though its year is 1973.
The full year of the ISO week number, as a decimal number.
Equivalent to ‘%b’.
The hour (24-hour clock) as a decimal number (00–23).
The hour (12-hour clock) as a decimal number (01–12).
The day of the year as a decimal number (001–366).
The month as a decimal number (01–12).
The minute as a decimal number (00–59).
A newline character (ASCII LF).
The locale's equivalent of the AM/PM designations associated with a 12-hour clock.
The locale's 12-hour clock time. (This is ‘%I:%M:%S %p’ in the "C" locale.)
Equivalent to specifying ‘%H:%M’.
The second as a decimal number (00–60).
A TAB character.
Equivalent to specifying ‘%H:%M:%S’.
The weekday as a decimal number (1–7). Monday is day one.
The week number of the year (the first Sunday as the first day of week one) as a decimal number (00–53).
The week number of the year (the first Monday as the first day of week one) as a decimal number (01–53). The method for determining the week number is as specified by ISO 8601. (To wit: if the week containing January 1 has four or more days in the new year, then it is week one; otherwise it is week 53 of the previous year and the next week is week one.)
The weekday as a decimal number (0–6). Sunday is day zero.
The week number of the year (the first Monday as the first day of week one) as a decimal number (00–53).
The locale's “appropriate” date representation. (This is ‘%A %B %d %Y’ in the "C" locale.)
The locale's “appropriate” time representation. (This is ‘%T’ in the "C" locale.)
The year modulo 100 as a decimal number (00–99).
The full year as a decimal number (e.g., 1995).
The timezone offset in a +HHMM format (e.g., the format necessary to produce RFC 822/RFC 1036 date headers).
The time zone name or abbreviation; no characters if no time zone is determinable.
%Ec %EC %Ex %EX %Ey %EY %Od %Oe %OH
%OI %Om %OM %OS %Ou %OU %OV %Ow %OW %Oy
“Alternate representations” for the specifications that use only the second letter (‘%c’, ‘%C’, and so on).43 (These facilitate compliance with the POSIX date utility.)
A literal ‘%’.

If a conversion specifier is not one of the above, the behavior is undefined.44

Informally, a locale is the geographic place in which a program is meant to run. For example, a common way to abbreviate the date September 4, 1991 in the United States is “9/4/91.” In many countries in Europe, however, it is abbreviated “4.9.91.” Thus, the ‘%x’ specification in a "US" locale might produce ‘9/4/91’, while in a "EUROPE" locale, it might produce ‘4.9.91’. The ISO C standard defines a default "C" locale, which is an environment that is typical of what most C programmers are used to.

For systems that are not yet fully standards-compliant, gawk supplies a copy of strftime from the GNU C Library. It supports all of the just listed format specifications. If that version is used to compile gawk (see Installation), then the following additional format specifications are available:

The hour (24-hour clock) as a decimal number (0–23). Single-digit numbers are padded with a space.
The hour (12-hour clock) as a decimal number (1–12). Single-digit numbers are padded with a space.
The time as a decimal timestamp in seconds since the epoch.

Additionally, the alternate representations are recognized but their normal representations are used.

This example is an awk implementation of the POSIX date utility. Normally, the date utility prints the current date and time of day in a well-known format. However, if you provide an argument to it that begins with a ‘+’, date copies nonformat specifier characters to the standard output and interprets the current time according to the format specifiers in the string. For example:

     $ date '+Today is %A, %B %d, %Y.'
     -| Today is Thursday, September 14, 2000.

Here is the gawk version of the date utility. It has a shell “wrapper” to handle the -u option, which requires that date run as if the time zone is set to UTC:

     #! /bin/sh
     # date --- approximate the P1003.2 'date' command
     case $1 in
     -u)  TZ=UTC0     # use UTC
          export TZ
          shift ;;
     gawk 'BEGIN  {
         format = "%a %b %d %H:%M:%S %Z %Y"
         exitval = 0
         if (ARGC > 2)
             exitval = 1
         else if (ARGC == 2) {
             format = ARGV[1]
             if (format ~ /^\+/)
                 format = substr(format, 2)   # remove leading +
         print strftime(format)
         exit exitval
     }' "$@"

Next: , Previous: Time Functions, Up: Built-in

8.1.6 Bit-Manipulation Functions of gawk

I can explain it for you, but I can't understand it for you.

Many languages provide the ability to perform bitwise operations on two integer numbers. In other words, the operation is performed on each successive pair of bits in the operands. Three common operations are bitwise AND, OR, and XOR. The operations are described in table-bitwise-ops.

                     Bit Operator
               |  AND  |   OR  |  XOR
     Operands  | 0 | 1 | 0 | 1 | 0 | 1
         0     | 0   0 | 0   1 | 0   1
         1     | 0   1 | 1   1 | 1   0

Table 8.6: Bitwise Operations

As you can see, the result of an AND operation is 1 only when both bits are 1. The result of an OR operation is 1 if either bit is 1. The result of an XOR operation is 1 if either bit is 1, but not both. The next operation is the complement; the complement of 1 is 0 and the complement of 0 is 1. Thus, this operation “flips” all the bits of a given value.

Finally, two other common operations are to shift the bits left or right. For example, if you have a bit string ‘10111001’ and you shift it right by three bits, you end up with ‘00010111’.45 If you start over again with ‘10111001’ and shift it left by three bits, you end up with ‘11001000’. gawk provides built-in functions that implement the bitwise operations just described. They are:

and(v1, v2) Returns the bitwise AND of the values provided by v1 and v2.

or(v1, v2) Returns the bitwise OR of the values provided by v1 and v2.

xor(v1, v2) Returns the bitwise XOR of the values provided by v1 and v2.

compl(val) Returns the bitwise complement of val.

lshift(val, count) Returns the value of val, shifted left by count bits.

rshift(val, count) Returns the value of val, shifted right by count bits.

For all of these functions, first the double-precision floating-point value is converted to the widest C unsigned integer type, then the bitwise operation is performed. If the result cannot be represented exactly as a C double, leading nonzero bits are removed one by one until it can be represented exactly. The result is then converted back into a C double. (If you don't understand this paragraph, don't worry about it.)

Here is a user-defined function (see User-defined) that illustrates the use of these functions:

     # bits2str --- turn a byte into readable 1's and 0's
     function bits2str(bits,        data, mask)
         if (bits == 0)
             return "0"
         mask = 1
         for (; bits != 0; bits = rshift(bits, 1))
             data = (and(bits, mask) ? "1" : "0") data
         while ((length(data) % 8) != 0)
             data = "0" data
         return data
     BEGIN {
         printf "123 = %s\n", bits2str(123)
         printf "0123 = %s\n", bits2str(0123)
         printf "0x99 = %s\n", bits2str(0x99)
         comp = compl(0x99)
         printf "compl(0x99) = %#x = %s\n", comp, bits2str(comp)
         shift = lshift(0x99, 2)
         printf "lshift(0x99, 2) = %#x = %s\n", shift, bits2str(shift)
         shift = rshift(0x99, 2)
         printf "rshift(0x99, 2) = %#x = %s\n", shift, bits2str(shift)

This program produces the following output when run:

     $ gawk -f testbits.awk
     -| 123 = 01111011
     -| 0123 = 01010011
     -| 0x99 = 10011001
     -| compl(0x99) = 0xffffff66 = 11111111111111111111111101100110
     -| lshift(0x99, 2) = 0x264 = 0000001001100100
     -| rshift(0x99, 2) = 0x26 = 00100110

The bits2str function turns a binary number into a string. The number 1 represents a binary value where the rightmost bit is set to 1. Using this mask, the function repeatedly checks the rightmost bit. ANDing the mask with the value indicates whether the rightmost bit is 1 or not. If so, a "1" is concatenated onto the front of the string. Otherwise, a "0" is added. The value is then shifted right by one bit and the loop continues until there are no more 1 bits.

If the initial value is zero it returns a simple "0". Otherwise, at the end, it pads the value with zeros to represent multiples of 8-bit quantities. This is typical in modern computers.

The main code in the BEGIN rule shows the difference between the decimal and octal values for the same numbers (see Nondecimal-numbers), and then demonstrates the results of the compl, lshift, and rshift functions.

Previous: Bitwise Functions, Up: Built-in

8.1.7 Using gawk's String-Translation Functions

gawk provides facilities for internationalizing awk programs. These include the functions described in the following list. The descriptions here are purposely brief. See Internationalization, for the full story. Optional parameters are enclosed in square brackets ([ ]):

dcgettext(string [, domain [, category]])
This function returns the translation of string in text domain domain for locale category category. The default value for domain is the current value of TEXTDOMAIN. The default value for category is "LC_MESSAGES".

dcngettext(string1, string2, number [, domain [, category]])
This function returns the plural form used for number of the translation of string1 and string2 in text domain domain for locale category category. string1 is the English singular variant of a message, and string2 the English plural variant of the same message. The default value for domain is the current value of TEXTDOMAIN. The default value for category is "LC_MESSAGES".

bindtextdomain(directory [, domain])
This function allows you to specify the directory in which gawk will look for message translation files, in case they will not or cannot be placed in the “standard” locations (e.g., during testing). It returns the directory in which domain is “bound.”

The default domain is the value of TEXTDOMAIN. If directory is the null string (""), then bindtextdomain returns the current binding for the given domain.

Previous: Built-in, Up: Functions

8.2 User-Defined Functions

Complicated awk programs can often be simplified by defining your own functions. User-defined functions can be called just like built-in ones (see Function Calls), but it is up to you to define them, i.e., to tell awk what they should do.

Next: , Up: User-defined

8.2.1 Function Definition Syntax

Definitions of functions can appear anywhere between the rules of an awk program. Thus, the general form of an awk program is extended to include sequences of rules and user-defined function definitions. There is no need to put the definition of a function before all uses of the function. This is because awk reads the entire program before starting to execute any of it.

The definition of a function named name looks like this:

     function name(parameter-list)

name is the name of the function to define. A valid function name is like a valid variable name: a sequence of letters, digits, and underscores that doesn't start with a digit. Within a single awk program, any particular name can only be used as a variable, array, or function.

parameter-list is a list of the function's arguments and local variable names, separated by commas. When the function is called, the argument names are used to hold the argument values given in the call. The local variables are initialized to the empty string. A function cannot have two parameters with the same name, nor may it have a parameter with the same name as the function itself.

The body-of-function consists of awk statements. It is the most important part of the definition, because it says what the function should actually do. The argument names exist to give the body a way to talk about the arguments; local variables exist to give the body places to keep temporary values.

Argument names are not distinguished syntactically from local variable names. Instead, the number of arguments supplied when the function is called determines how many argument variables there are. Thus, if three argument values are given, the first three names in parameter-list are arguments and the rest are local variables.

It follows that if the number of arguments is not the same in all calls to the function, some of the names in parameter-list may be arguments on some occasions and local variables on others. Another way to think of this is that omitted arguments default to the null string.

Usually when you write a function, you know how many names you intend to use for arguments and how many you intend to use as local variables. It is conventional to place some extra space between the arguments and the local variables, in order to document how your function is supposed to be used.

During execution of the function body, the arguments and local variable values hide, or shadow, any variables of the same names used in the rest of the program. The shadowed variables are not accessible in the function definition, because there is no way to name them while their names have been taken away for the local variables. All other variables used in the awk program can be referenced or set normally in the function's body.

The arguments and local variables last only as long as the function body is executing. Once the body finishes, you can once again access the variables that were shadowed while the function was running.

The function body can contain expressions that call functions. They can even call this function, either directly or by way of another function. When this happens, we say the function is recursive. The act of a function calling itself is called recursion.

In many awk implementations, including gawk, the keyword function may be abbreviated func. However, POSIX only specifies the use of the keyword function. This actually has some practical implications. If gawk is in POSIX-compatibility mode (see Options), then the following statement does not define a function:

     func foo() { a = sqrt($1) ; print a }

Instead it defines a rule that, for each record, concatenates the value of the variable ‘func’ with the return value of the function ‘foo’. If the resulting string is non-null, the action is executed. This is probably not what is desired. (awk accepts this input as syntactically valid, because functions may be used before they are defined in awk programs.)

To ensure that your awk programs are portable, always use the keyword function when defining a function.

Next: , Previous: Definition Syntax, Up: User-defined

8.2.2 Function Definition Examples

Here is an example of a user-defined function, called myprint, that takes a number and prints it in a specific format:

     function myprint(num)
          printf "%6.3g\n", num

To illustrate, here is an awk rule that uses our myprint function:

     $3 > 0     { myprint($3) }

This program prints, in our special format, all the third fields that contain a positive number in our input. Therefore, when given the following:

      1.2   3.4    5.6   7.8
      9.10 11.12 -13.14 15.16
     17.18 19.20  21.22 23.24

this program, using our function to format the results, prints:


This function deletes all the elements in an array:

     function delarray(a,    i)
         for (i in a)
            delete a[i]

When working with arrays, it is often necessary to delete all the elements in an array and start over with a new list of elements (see Delete). Instead of having to repeat this loop everywhere that you need to clear out an array, your program can just call delarray. (This guarantees portability. The use of ‘delete array’ to delete the contents of an entire array is a nonstandard extension.)

The following is an example of a recursive function. It takes a string as an input parameter and returns the string in backwards order. Recursive functions must always have a test that stops the recursion. In this case, the recursion terminates when the starting position is zero, i.e., when there are no more characters left in the string.

     function rev(str, start)
         if (start == 0)
             return ""
         return (substr(str, start, 1) rev(str, start - 1))

If this function is in a file named rev.awk, it can be tested this way:

     $ echo "Don't Panic!" |
     > gawk --source '{ print rev($0, length($0)) }' -f rev.awk
     -| !cinaP t'noD

The C ctime function takes a timestamp and returns it in a string, formatted in a well-known fashion. The following example uses the built-in strftime function (see Time Functions) to create an awk version of ctime:

     # ctime.awk
     # awk version of C ctime(3) function
     function ctime(ts,    format)
         format = "%a %b %d %H:%M:%S %Z %Y"
         if (ts == 0)
             ts = systime()       # use current time as default
         return strftime(format, ts)

Next: , Previous: Function Example, Up: User-defined

8.2.3 Calling User-Defined Functions

Calling a function means causing the function to run and do its job. A function call is an expression and its value is the value returned by the function.

A function call consists of the function name followed by the arguments in parentheses. awk expressions are what you write in the call for the arguments. Each time the call is executed, these expressions are evaluated, and the values are the actual arguments. For example, here is a call to foo with three arguments (the first being a string concatenation):

     foo(x y, "lose", 4 * z)

Caution: Whitespace characters (spaces and tabs) are not allowed between the function name and the open-parenthesis of the argument list. If you write whitespace by mistake, awk might think that you mean to concatenate a variable with an expression in parentheses. However, it notices that you used a function name and not a variable name, and reports an error.

When a function is called, it is given a copy of the values of its arguments. This is known as call by value. The caller may use a variable as the expression for the argument, but the called function does not know this—it only knows what value the argument had. For example, if you write the following code:

     foo = "bar"
     z = myfunc(foo)

then you should not think of the argument to myfunc as being “the variable foo.” Instead, think of the argument as the string value "bar". If the function myfunc alters the values of its local variables, this has no effect on any other variables. Thus, if myfunc does this:

     function myfunc(str)
       print str
       str = "zzz"
       print str

to change its first argument variable str, it does not change the value of foo in the caller. The role of foo in calling myfunc ended when its value ("bar") was computed. If str also exists outside of myfunc, the function body cannot alter this outer value, because it is shadowed during the execution of myfunc and cannot be seen or changed from there.

However, when arrays are the parameters to functions, they are not copied. Instead, the array itself is made available for direct manipulation by the function. This is usually called call by reference. Changes made to an array parameter inside the body of a function are visible outside that function.

NOTE: Changing an array parameter inside a function can be very dangerous if you do not watch what you are doing. For example:
     function changeit(array, ind, nvalue)
          array[ind] = nvalue
     BEGIN {
         a[1] = 1; a[2] = 2; a[3] = 3
         changeit(a, 2, "two")
         printf "a[1] = %s, a[2] = %s, a[3] = %s\n",
                 a[1], a[2], a[3]

prints ‘a[1] = 1, a[2] = two, a[3] = 3’, because changeit stores "two" in the second element of a.

Some awk implementations allow you to call a function that has not been defined. They only report a problem at runtime when the program actually tries to call the function. For example:

     BEGIN {
         if (0)
     function bar() { ... }
     # note that `foo' is not defined

Because the ‘if’ statement will never be true, it is not really a problem that foo has not been defined. Usually, though, it is a problem if a program calls an undefined function.

If --lint is specified (see Options), gawk reports calls to undefined functions.

Some awk implementations generate a runtime error if you use the next statement (see Next Statement) inside a user-defined function. gawk does not have this limitation.

Next: , Previous: Function Caveats, Up: User-defined

8.2.4 The return Statement

The body of a user-defined function can contain a return statement. This statement returns control to the calling part of the awk program. It can also be used to return a value for use in the rest of the awk program. It looks like this:

     return [expression]

The expression part is optional. If it is omitted, then the returned value is undefined, and therefore, unpredictable.

A return statement with no value expression is assumed at the end of every function definition. So if control reaches the end of the function body, then the function returns an unpredictable value. awk does not warn you if you use the return value of such a function.

Sometimes, you want to write a function for what it does, not for what it returns. Such a function corresponds to a void function in C or to a procedure in Pascal. Thus, it may be appropriate to not return any value; simply bear in mind that if you use the return value of such a function, you do so at your own risk.

The following is an example of a user-defined function that returns a value for the largest number among the elements of an array:

     function maxelt(vec,   i, ret)
          for (i in vec) {
               if (ret == "" || vec[i] > ret)
                    ret = vec[i]
          return ret

You call maxelt with one argument, which is an array name. The local variables i and ret are not intended to be arguments; while there is nothing to stop you from passing more than one argument to maxelt, the results would be strange. The extra space before i in the function parameter list indicates that i and ret are not supposed to be arguments. You should follow this convention when defining functions.

The following program uses the maxelt function. It loads an array, calls maxelt, and then reports the maximum number in that array:

     function maxelt(vec,   i, ret)
          for (i in vec) {
               if (ret == "" || vec[i] > ret)
                    ret = vec[i]
          return ret
     # Load all fields of each record into nums.
          for(i = 1; i <= NF; i++)
               nums[NR, i] = $i
     END {
          print maxelt(nums)

Given the following input:

      1 5 23 8 16
     44 3 5 2 8 26
     256 291 1396 2962 100
     -6 467 998 1101
     99385 11 0 225

the program reports (predictably) that 99385 is the largest number in the array.

Previous: Return Statement, Up: User-defined

8.2.5 Functions and Their Effects on Variable Typing

awk is a very fluid language. It is possible that awk can't tell if an identifier represents a regular variable or an array until runtime. Here is an annotated sample program:

     function foo(a)
         a[1] = 1   # parameter is an array
     BEGIN {
         b = 1
         foo(b)  # invalid: fatal type mismatch
         foo(x)  # x uninitialized, becomes an array dynamically
         x = 1   # now not allowed, runtime error

Usually, such things aren't a big issue, but it's worth being aware of them.

Next: , Previous: Functions, Up: Top

9 Internationalization with gawk

Once upon a time, computer makers wrote software that worked only in English. Eventually, hardware and software vendors noticed that if their systems worked in the native languages of non-English-speaking countries, they were able to sell more systems. As a result, internationalization and localization of programs and software systems became a common practice.

Until recently, the ability to provide internationalization was largely restricted to programs written in C and C++. This chapter describes the underlying library gawk uses for internationalization, as well as how gawk makes internationalization features available at the awk program level. Having internationalization available at the awk level gives software developers additional flexibility—they are no longer required to write in C when internationalization is a requirement.

Next: , Up: Internationalization

9.1 Internationalization and Localization

Internationalization means writing (or modifying) a program once, in such a way that it can use multiple languages without requiring further source-code changes. Localization means providing the data necessary for an internationalized program to work in a particular language. Most typically, these terms refer to features such as the language used for printing error messages, the language used to read responses, and information related to how numerical and monetary values are printed and read.

Next: , Previous: I18N and L10N, Up: Internationalization

9.2 GNU gettext

The facilities in GNU gettext focus on messages; strings printed by a program, either directly or via formatting with printf or sprintf.46

When using GNU gettext, each application has its own text domain. This is a unique name, such as ‘kpilot’ or ‘gawk’, that identifies the application. A complete application may have multiple components—programs written in C or C++, as well as scripts written in sh or awk. All of the components use the same text domain.

To make the discussion concrete, assume we're writing an application named guide. Internationalization consists of the following steps, in this order:

  1. The programmer goes through the source for all of guide's components and marks each string that is a candidate for translation. For example, "`-F': option required" is a good candidate for translation. A table with strings of option names is not (e.g., gawk's --profile option should remain the same, no matter what the local language).

  2. The programmer indicates the application's text domain ("guide") to the gettext library, by calling the textdomain function.
  3. Messages from the application are extracted from the source code and collected into a portable object file (guide.po), which lists the strings and their translations. The translations are initially empty. The original (usually English) messages serve as the key for lookup of the translations.

  4. For each language with a translator, guide.po is copied and translations are created and shipped with the application.

  5. Each language's .po file is converted into a binary message object (.mo) file. A message object file contains the original messages and their translations in a binary format that allows fast lookup of translations at runtime.
  6. When guide is built and installed, the binary translation files are installed in a standard place.

  7. For testing and development, it is possible to tell gettext to use .mo files in a different directory than the standard one by using the bindtextdomain function.

  8. At runtime, guide looks up each string via a call to gettext. The returned string is the translated string if available, or the original string if not.
  9. If necessary, it is possible to access messages from a different text domain than the one belonging to the application, without having to switch the application's default text domain back and forth.

In C (or C++), the string marking and dynamic translation lookup are accomplished by wrapping each string in a call to gettext:

     printf(gettext("Don't Panic!\n"));

The tools that extract messages from source code pull out all strings enclosed in calls to gettext.

The GNU gettext developers, recognizing that typing ‘gettext’ over and over again is both painful and ugly to look at, use the macro ‘_’ (an underscore) to make things easier:

     /* In the standard header file: */
     #define _(str) gettext(str)
     /* In the program text: */
     printf(_("Don't Panic!\n"));

This reduces the typing overhead to just three extra characters per string and is considerably easier to read as well. There are locale categories for different types of locale-related information. The defined locale categories that gettext knows about are:

Text messages. This is the default category for gettext operations, but it is possible to supply a different one explicitly, if necessary. (It is almost never necessary to supply a different category.)

Text-collation information; i.e., how different characters and/or groups of characters sort in a given language.

Character-type information (alphabetic, digit, upper- or lowercase, and so on). This information is accessed via the POSIX character classes in regular expressions, such as /[[:alnum:]]/ (see Regexp Operators).

Monetary information, such as the currency symbol, and whether the symbol goes before or after a number.

Numeric information, such as which characters to use for the decimal point and the thousands separator.47

Response information, such as how “yes” and “no” appear in the local language, and possibly other information as well.

Time- and date-related information, such as 12- or 24-hour clock, month printed before or after day in a date, local month abbreviations, and so on.

All of the above. (Not too useful in the context of gettext.)

Next: , Previous: Explaining gettext, Up: Internationalization

9.3 Internationalizing awk Programs

gawk provides the following variables and functions for internationalization:

This variable indicates the application's text domain. For compatibility with GNU gettext, the default value is "messages".

_"your message here"
String constants marked with a leading underscore are candidates for translation at runtime. String constants without a leading underscore are not translated.

dcgettext(string [, domain [, category]])
This built-in function returns the translation of string in text domain domain for locale category category. The default value for domain is the current value of TEXTDOMAIN. The default value for category is "LC_MESSAGES".

If you supply a value for category, it must be a string equal to one of the known locale categories described in the previous section. You must also supply a text domain. Use TEXTDOMAIN if you want to use the current domain.

Caution: The order of arguments to the awk version of the dcgettext function is purposely different from the order for the C version. The awk version's order was chosen to be simple and to allow for reasonable awk-style default arguments.

dcngettext(string1, string2, number [, domain [, category]])
This built-in function returns the plural form used for number of the translation of string1 and string2 in text domain domain for locale category category. string1 is the English singular variant of a message, and string2 the English plural variant of the same message. The default value for domain is the current value of TEXTDOMAIN. The default value for category is "LC_MESSAGES".

The same remarks as for the dcgettext function apply.

bindtextdomain(directory [, domain])
This built-in function allows you to specify the directory in which gettext looks for .mo files, in case they will not or cannot be placed in the standard locations (e.g., during testing). It returns the directory in which domain is “bound.”

The default domain is the value of TEXTDOMAIN. If directory is the null string (""), then bindtextdomain returns the current binding for the given domain.

To use these facilities in your awk program, follow the steps outlined in the previous section, like so:

  1. Set the variable TEXTDOMAIN to the text domain of your program. This is best done in a BEGIN rule (see BEGIN/END), or it can also be done via the -v command-line option (see Options):
              BEGIN {
                  TEXTDOMAIN = "guide"

  2. Mark all translatable strings with a leading underscore (‘_’) character. It must be adjacent to the opening quote of the string. For example:
              print _"hello, world"
              x = _"you goofed"
              printf(_"Number of users is %d\n", nusers)
  3. If you are creating strings dynamically, you can still translate them, using the dcgettext built-in function:
              message = nusers " users logged in"
              message = dcgettext(message, "adminprog")
              print message

    Here, the call to dcgettext supplies a different text domain ("adminprog") in which to find the message, but it uses the default "LC_MESSAGES" category.

  4. During development, you might want to put the .mo file in a private directory for testing. This is done with the bindtextdomain built-in function:
              BEGIN {
                 TEXTDOMAIN = "guide"   # our text domain
                 if (Testing) {
                     # where to find our files
                     # joe is in charge of adminprog
                     bindtextdomain("../joe/testdir", "adminprog")

See I18N Example, for an example program showing the steps to create and use translations from awk.

Next: , Previous: Programmer i18n, Up: Internationalization

9.4 Translating awk Programs

Once a program's translatable strings have been marked, they must be extracted to create the initial .po file. As part of translation, it is often helpful to rearrange the order in which arguments to printf are output.

gawk's --gen-po command-line option extracts the messages and is discussed next. After that, printf's ability to rearrange the order for printf arguments at runtime is covered.

Next: , Up: Translator i18n

9.4.1 Extracting Marked Strings

Once your awk program is working, and all the strings have been marked and you've set (and perhaps bound) the text domain, it is time to produce translations. First, use the --gen-po command-line option to create the initial .po file:

     $ gawk --gen-po -f guide.awk > guide.po

When run with --gen-po, gawk does not execute your program. Instead, it parses it as usual and prints all marked strings to standard output in the format of a GNU gettext Portable Object file. Also included in the output are any constant strings that appear as the first argument to dcgettext or as the first and second argument to dcngettext.48 See I18N Example, for the full list of steps to go through to create and test translations for guide.

Next: , Previous: String Extraction, Up: Translator i18n

9.4.2 Rearranging printf Arguments

Format strings for printf and sprintf (see Printf) present a special problem for translation. Consider the following:49

     printf(_"String `%s' has %d characters\n",
               string, length(string)))

A possible German translation for this might be:

     "%d Zeichen lang ist die Zeichenkette `%s'\n"

The problem should be obvious: the order of the format specifications is different from the original! Even though gettext can return the translated string at runtime, it cannot change the argument order in the call to printf.

To solve this problem, printf format specifiers may have an additional optional element, which we call a positional specifier. For example:

     "%2$d Zeichen lang ist die Zeichenkette `%1$s'\n"

Here, the positional specifier consists of an integer count, which indicates which argument to use, and a ‘$’. Counts are one-based, and the format string itself is not included. Thus, in the following example, ‘string’ is the first argument and ‘length(string)’ is the second:

     $ gawk 'BEGIN {
     >     string = "Dont Panic"
     >     printf _"%2$d characters live in \"%1$s\"\n",
     >                         string, length(string)
     > }'
     -| 10 characters live in "Dont Panic"

If present, positional specifiers come first in the format specification, before the flags, the field width, and/or the precision.

Positional specifiers can be used with the dynamic field width and precision capability:

     $ gawk 'BEGIN {
     >    printf("%*.*s\n", 10, 20, "hello")
     >    printf("%3$*2$.*1$s\n", 20, 10, "hello")
     > }'
     -|      hello
     -|      hello
NOTE: When using ‘*’ with a positional specifier, the ‘*’ comes first, then the integer position, and then the ‘$’. This is somewhat counterintuitive.

gawk does not allow you to mix regular format specifiers and those with positional specifiers in the same string:

     $ gawk 'BEGIN { printf _"%d %3$s\n", 1, 2, "hi" }'
     error--> gawk: cmd. line:1: fatal: must use `count$' on all formats or none
NOTE: There are some pathological cases that gawk may fail to diagnose. In such cases, the output may not be what you expect. It's still a bad idea to try mixing them, even if gawk doesn't detect it.

Although positional specifiers can be used directly in awk programs, their primary purpose is to help in producing correct translations of format strings into languages different from the one in which the program is first written.

Previous: Printf Ordering, Up: Translator i18n

9.4.3 awk Portability Issues

gawk's internationalization features were purposely chosen to have as little impact as possible on the portability of awk programs that use them to other versions of awk. Consider this program:

     BEGIN {
         TEXTDOMAIN = "guide"
         if (Test_Guide)   # set with -v
         print _"don't panic!"

As written, it won't work on other versions of awk. However, it is actually almost portable, requiring very little change:

Next: , Previous: Translator i18n, Up: Internationalization

9.5 A Simple Internationalization Example

Now let's look at a step-by-step example of how to internationalize and localize a simple awk program, using guide.awk as our original source:

     BEGIN {
         TEXTDOMAIN = "guide"
         bindtextdomain(".")  # for testing
         print _"Don't Panic"
         print _"The Answer Is", 42
         print "Pardon me, Zaphod who?"

Run ‘gawk --gen-po’ to create the .po file:

     $ gawk --gen-po -f guide.awk > guide.po

This produces:

     #: guide.awk:4
     msgid "Don't Panic"
     msgstr ""
     #: guide.awk:5
     msgid "The Answer Is"
     msgstr ""

This original portable object file is saved and reused for each language into which the application is translated. The msgid is the original string and the msgstr is the translation.

NOTE: Strings not marked with a leading underscore do not appear in the guide.po file.

Next, the messages must be translated. Here is a translation to a hypothetical dialect of English, called “Mellow”:51

     $ cp guide.po guide-mellow.po
     Add translations to guide-mellow.po ...

Following are the translations:

     #: guide.awk:4
     msgid "Don't Panic"
     msgstr "Hey man, relax!"
     #: guide.awk:5
     msgid "The Answer Is"
     msgstr "Like, the scoop is"

The next step is to make the directory to hold the binary message object file and then to create the file. The directory layout shown here is standard for GNU gettext on GNU/Linux systems. Other versions of gettext may use a different layout:

     $ mkdir en_US en_US/LC_MESSAGES

The msgfmt utility does the conversion from human-readable .po file to machine-readable .mo file. By default, msgfmt creates a file named messages. This file must be renamed and placed in the proper directory so that gawk can find it:

     $ msgfmt guide-mellow.po
     $ mv messages en_US/LC_MESSAGES/

Finally, we run the program to test it:

     $ gawk -f guide.awk
     -| Hey man, relax!
     -| Like, the scoop is 42
     -| Pardon me, Zaphod who?

If the three replacement functions for dcgettext, dcngettext and bindtextdomain (see I18N Portability) are in a file named libintl.awk, then we can run guide.awk unchanged as follows:

     $ gawk --posix -f guide.awk -f libintl.awk
     -| Don't Panic
     -| The Answer Is 42
     -| Pardon me, Zaphod who?

Previous: I18N Example, Up: Internationalization

9.6 gawk Can Speak Your Language

As of version 3.1, gawk itself has been internationalized using the GNU gettext package. (GNU gettext is described in complete detail in GNU gettext tools.) As of this writing, the latest version of GNU gettext is version 0.11.5.

If a translation of gawk's messages exists, then gawk produces usage messages, warnings, and fatal errors in the local language.

Next: , Previous: Internationalization, Up: Top

10 Advanced Features of gawk

Write documentation as if whoever reads it is a violent psychopath who knows where you live.
Steve English, as quoted by Peter Langston

This chapter discusses advanced features in gawk. It's a bit of a “grab bag” of items that are otherwise unrelated to each other. First, a command-line option allows gawk to recognize nondecimal numbers in input data, not just in awk programs. Next, two-way I/O, discussed briefly in earlier parts of this Web page, is described in full detail, along with the basics of TCP/IP networking and BSD portal files. Finally, gawk can profile an awk program, making it possible to tune it for performance.

Dynamic Extensions, discusses the ability to dynamically add new built-in functions to gawk. As this feature is still immature and likely to change, its description is relegated to an appendix.

Next: , Up: Advanced Features

10.1 Allowing Nondecimal Input Data

If you run gawk with the --non-decimal-data option, you can have nondecimal constants in your input data:

     $ echo 0123 123 0x123 |
     > gawk --non-decimal-data '{ printf "%d, %d, %d\n",
     >                                         $1, $2, $3 }'
     -| 83, 123, 291

For this feature to work, write your program so that gawk treats your data as numeric:

     $ echo 0123 123 0x123 | gawk '{ print $1, $2, $3 }'
     -| 0123 123 0x123

The print statement treats its expressions as strings. Although the fields can act as numbers when necessary, they are still strings, so print does not try to treat them numerically. You may need to add zero to a field to force it to be treated as a number. For example:

     $ echo 0123 123 0x123 | gawk --non-decimal-data '
     > { print $1, $2, $3
     >   print $1 + 0, $2 + 0, $3 + 0 }'
     -| 0123 123 0x123
     -| 83 123 291

Because it is common to have decimal data with leading zeros, and because using it could lead to surprising results, the default is to leave this facility disabled. If you want it, you must explicitly request it.

Caution: Use of this option is not recommended. It can break old programs very badly. Instead, use the strtonum function to convert your data (see Nondecimal-numbers). This makes your programs easier to write and easier to read, and leads to less surprising results.

Next: , Previous: Nondecimal Data, Up: Advanced Features

10.2 Two-Way Communications with Another Process

     From: (Mike Brennan)
     Newsgroups: comp.lang.awk
     Subject: Re: Learn the SECRET to Attract Women Easily
     Date: 4 Aug 1997 17:34:46 GMT
     Message-ID: <5s53rm$>
     On 3 Aug 1997 13:17:43 GMT, Want More Dates???
     <> wrote:
     >Learn the SECRET to Attract Women Easily
     >The SCENT(tm)  Pheromone Sex Attractant For Men to Attract Women
     The scent of awk programmers is a lot more attractive to women than
     the scent of perl programmers.
     Mike Brennan

It is often useful to be able to send data to a separate program for processing and then read the result. This can always be done with temporary files:

     # write the data for processing
     tempfile = ("mydata." PROCINFO["pid"])
     while (not done with data)
         print data | ("subprogram > " tempfile)
     close("subprogram > " tempfile)
     # read the results, remove tempfile when done
     while ((getline newdata < tempfile) > 0)
         process newdata appropriately
     system("rm " tempfile)

This works, but not elegantly. Among other things, it requires that the program be run in a directory that cannot be shared among users; for example, /tmp will not do, as another user might happen to be using a temporary file with the same name.

Starting with version 3.1 of gawk, it is possible to open a two-way pipe to another process. The second process is termed a coprocess, since it runs in parallel with gawk. The two-way connection is created using the new ‘|&’ operator (borrowed from the Korn shell, ksh):52

     do {
         print data |& "subprogram"
         "subprogram" |& getline results
     } while (data left to process)

The first time an I/O operation is executed using the ‘|&’ operator, gawk creates a two-way pipeline to a child process that runs the other program. Output created with print or printf is written to the program's standard input, and output from the program's standard output can be read by the gawk program using getline. As is the case with processes started by ‘|’, the subprogram can be any program, or pipeline of programs, that can be started by the shell.

There are some cautionary items to be aware of:

It is possible to close just one end of the two-way pipe to a coprocess, by supplying a second argument to the close function of either "to" or "from" (see Close Files And Pipes). These strings tell gawk to close the end of the pipe that sends data to the process or the end that reads from it, respectively.

This is particularly necessary in order to use the system sort utility as part of a coprocess; sort must read all of its input data before it can produce any output. The sort program does not receive an end-of-file indication until gawk closes the write end of the pipe.

When you have finished writing data to the sort utility, you can close the "to" end of the pipe, and then start reading sorted data via getline. For example:

     BEGIN {
         command = "LC_ALL=C sort"
         n = split("abcdefghijklmnopqrstuvwxyz", a, "")
         for (i = n; i > 0; i--)
             print a[i] |& command
         close(command, "to")
         while ((command |& getline line) > 0)
             print "got", line

This program writes the letters of the alphabet in reverse order, one per line, down the two-way pipe to sort. It then closes the write end of the pipe, so that sort receives an end-of-file indication. This causes sort to sort the data and write the sorted data back to the gawk program. Once all of the data has been read, gawk terminates the coprocess and exits.

As a side note, the assignment ‘LC_ALL=C’ in the sort command ensures traditional Unix (ASCII) sorting from sort.

Beginning with gawk 3.1.2, you may use Pseudo-ttys (ptys) for two-way communication instead of pipes, if your system supports them. This is done on a per-command basis, by setting a special element in the PROCINFO array (see Auto-set), like so:

     command = "sort -nr"           # command, saved in variable for convenience
     PROCINFO[command, "pty"] = 1   # update PROCINFO
     print ... |& command       # start two-way pipe

Using ptys avoids the buffer deadlock issues described earlier, at some loss in performance. If your system does not have ptys, or if all the system's ptys are in use, gawk automatically falls back to using regular pipes.

Next: , Previous: Two-way I/O, Up: Advanced Features

10.3 Using gawk for Network Programming

EMISTERED: A host is a host from coast to coast,
and no-one can talk to host that's close,
unless the host that isn't close
is busy hung or dead.

In addition to being able to open a two-way pipeline to a coprocess on the same system (see Two-way I/O), it is possible to make a two-way connection to another process on another system across an IP networking connection.

You can think of this as just a very long two-way pipeline to a coprocess. The way gawk decides that you want to use TCP/IP networking is by recognizing special file names that begin with ‘/inet/’.

The full syntax of the special file name is /inet/protocol/local-port/remote-host/remote-port. The components are:

The protocol to use over IP. This must be either ‘tcp’, ‘udp’, or ‘raw’, for a TCP, UDP, or raw IP connection, respectively. The use of TCP is recommended for most applications.

Caution: The use of raw sockets is not currently supported in version 3.1 of gawk.

The local TCP or UDP port number to use. Use a port number of ‘0’ when you want the system to pick a port. This is what you should do when writing a TCP or UDP client. You may also use a well-known service name, such as ‘smtp’ or ‘http’, in which case gawk attempts to determine the predefined port number using the C getservbyname function.
The IP address or fully-qualified domain name of the Internet host to which you want to connect.
The TCP or UDP port number to use on the given remote-host. Again, use ‘0’ if you don't care, or else a well-known service name.

Consider the following very simple example:

     BEGIN {
       Service = "/inet/tcp/0/localhost/daytime"
       Service |& getline
       print $0

This program reads the current date and time from the local system's TCP ‘daytime’ server. It then prints the results and closes the connection.

Because this topic is extensive, the use of gawk for TCP/IP programming is documented separately. See TCP/IP Internetworking with gawk, which comes as part of the gawk distribution, for a much more complete introduction and discussion, as well as extensive examples.

Next: , Previous: TCP/IP Networking, Up: Advanced Features

10.4 Using gawk with BSD Portals

Similar to the /inet special files, if gawk is configured with the --enable-portals option (see Quick Installation), then gawk treats files whose pathnames begin with /p as 4.4 BSD-style portals.

When used with the ‘|&’ operator, gawk opens the file for two-way communications. The operating system's portal mechanism then manages creating the process associated with the portal and the corresponding communications with the portal's process.

Previous: Portal Files, Up: Advanced Features

10.5 Profiling Your awk Programs

Beginning with version 3.1 of gawk, you may produce execution traces of your awk programs. This is done with a specially compiled version of gawk, called pgawk (“profiling gawk”).

pgawk is identical in every way to gawk, except that when it has finished running, it creates a profile of your program in a file named awkprof.out. Because it is profiling, it also executes up to 45% slower than gawk normally does.

As shown in the following example, the --profile option can be used to change the name of the file where pgawk will write the profile:

     $ pgawk -f myprog.awk data1 data2

In the above example, pgawk places the profile in instead of in awkprof.out.

Regular gawk also accepts this option. When called with just --profile, gawk “pretty prints” the program into awkprof.out, without any execution counts. You may supply an option to --profile to change the file name. Here is a sample session showing a simple awk program, its input data, and the results from running pgawk. First, the awk program:

     BEGIN { print "First BEGIN rule" }
     END { print "First END rule" }
     /foo/ {
         print "matched /foo/, gosh"
         for (i = 1; i <= 3; i++)
         if (/foo/)
             print "if is true"
             print "else is true"
     BEGIN { print "Second BEGIN rule" }
     END { print "Second END rule" }
     function sing(    dummy)
         print "I gotta be me!"

Following is the input data:


Here is the awkprof.out that results from running pgawk on this program and data (this example also illustrates that awk programmers sometimes have to work late):

             # gawk profile, created Sun Aug 13 00:00:15 2000
             # BEGIN block(s)
             BEGIN {
          1          print "First BEGIN rule"
          1          print "Second BEGIN rule"
             # Rule(s)
          5  /foo/   { # 2
          2          print "matched /foo/, gosh"
          6          for (i = 1; i <= 3; i++) {
          6                  sing()
          5  {
          5          if (/foo/) { # 2
          2                  print "if is true"
          3          } else {
          3                  print "else is true"
             # END block(s)
             END {
          1          print "First END rule"
          1          print "Second END rule"
             # Functions, listed alphabetically
          6  function sing(dummy)
          6          print "I gotta be me!"

This example illustrates many of the basic rules for profiling output. The rules are as follows:

The profiled version of your program may not look exactly like what you typed when you wrote it. This is because pgawk creates the profiled version by “pretty printing” its internal representation of the program. The advantage to this is that pgawk can produce a standard representation. The disadvantage is that all source-code comments are lost, as are the distinctions among multiple BEGIN and END rules. Also, things such as:


come out as:

     /foo/   {
         print $0

which is correct, but possibly surprising.

Besides creating profiles when a program has completed, pgawk can produce a profile while it is running. This is useful if your awk program goes into an infinite loop and you want to see what has been executed. To use this feature, run pgawk in the background:

     $ pgawk -f myprog &
     [1] 13992

The shell prints a job number and process ID number; in this case, 13992. Use the kill command to send the USR1 signal to pgawk:

     $ kill -USR1 13992

As usual, the profiled version of the program is written to awkprof.out, or to a different file if you use the --profile option.

Along with the regular profile, as shown earlier, the profile includes a trace of any active functions:

     # Function Call Stack:
     #   3. baz
     #   2. bar
     #   1. foo
     # -- main --

You may send pgawk the USR1 signal as many times as you like. Each time, the profile and function call trace are appended to the output profile file.

If you use the HUP signal instead of the USR1 signal, pgawk produces the profile and the function call trace and then exits.

When pgawk runs on MS-DOS or MS-Windows, it uses the INT and QUIT signals for producing the profile and, in the case of the INT signal, pgawk exits. This is because these systems don't support the kill command, so the only signals you can deliver to a program are those generated by the keyboard. The INT signal is generated by the Ctrl-<C> or Ctrl-<BREAK> key, while the QUIT signal is generated by the Ctrl-<\> key.

Next: , Previous: Advanced Features, Up: Top

11 Running awk and gawk

This chapter covers how to run awk, both POSIX-standard and gawk-specific command-line options, and what awk and gawk do with non-option arguments. It then proceeds to cover how gawk searches for source files, obsolete options and/or features, and known bugs in gawk. This chapter rounds out the discussion of awk as a program and as a language.

While a number of the options and features described here were discussed in passing earlier in the book, this chapter provides the full details.

Next: , Up: Invoking Gawk

11.1 Invoking awk

There are two ways to run awk—with an explicit program or with one or more program files. Here are templates for both of them; items enclosed in [...] in these templates are optional:

     awk [options] -f progfile [--] file ...
     awk [options] [--] 'program' file ...

Besides traditional one-letter POSIX-style options, gawk also supports GNU long options.

It is possible to invoke awk with an empty program:

     awk '' datafile1 datafile2

Doing so makes little sense, though; awk exits silently when given an empty program. (d.c.) If --lint has been specified on the command line, gawk issues a warning that the program is empty.

Next: , Previous: Command Line, Up: Invoking Gawk

11.2 Command-Line Options

Options begin with a dash and consist of a single character. GNU-style long options consist of two dashes and a keyword. The keyword can be abbreviated, as long as the abbreviation allows the option to be uniquely identified. If the option takes an argument, then the keyword is either immediately followed by an equals sign (‘=’) and the argument's value, or the keyword and the argument's value are separated by whitespace. If a particular option with a value is given more than once, it is the last value that counts.

Each long option for gawk has a corresponding POSIX-style option. The long and short options are interchangeable in all contexts. The options and their meanings are as follows:

-F fs
--field-separator fs
Sets the FS variable to fs (see Field Separators).
-f source-file
--file source-file
Indicates that the awk program is to be found in source-file instead of in the first non-option argument.
-v var=val
--assign var=val
Sets the variable var to the value val before execution of the program begins. Such variable values are available inside the BEGIN rule (see Other Arguments).

The -v option can only set one variable, but it can be used more than once, setting another variable each time, like this: ‘awk -v foo=1 -v bar=2 ...’.

Caution: Using -v to set the values of the built-in variables may lead to surprising results. awk will reset the values of those variables as it needs to, possibly ignoring any predefined value you may have given.

-mf N
-mr N
Sets various memory limits to the value N. The ‘f’ flag sets the maximum number of fields and the ‘r’ flag sets the maximum record size. These two flags and the -m option are from the Bell Laboratories research version of Unix awk. They are provided for compatibility but otherwise ignored by gawk, since gawk has no predefined limits. (The Bell Laboratories awk no longer needs these options; it continues to accept them to avoid breaking old programs.)
-W gawk-opt
Following the POSIX standard, implementation-specific options are supplied as arguments to the -W option. These options also have corresponding GNU-style long options. Note that the long options may be abbreviated, as long as the abbreviations remain unique. The full list of gawk-specific options is provided next.
Signals the end of the command-line options. The following arguments are not treated as options even if they begin with ‘-’. This interpretation of -- follows the POSIX argument parsing conventions.

This is useful if you have file names that start with ‘-’, or in shell scripts, if you have file names that will be specified by the user that could start with ‘-’.

The previous list described options mandated by the POSIX standard, as well as options available in the Bell Laboratories version of awk. The following list describes gawk-specific options:

-W compat
-W traditional
Specifies compatibility mode, in which the GNU extensions to the awk language are disabled, so that gawk behaves just like the Bell Laboratories research version of Unix awk. --traditional is the preferred form of this option. See POSIX/GNU, which summarizes the extensions. Also see Compatibility Mode.
-W copyright
Print the short version of the General Public License and then exit.
-W copyleft
Just like --copyright. This option may disappear in a future version of gawk.

-W dump-variables[=file]
Prints a sorted list of global variables, their types, and final values to file. If no file is provided, gawk prints this list to the file named awkvars.out in the current directory.

Having a list of all global variables is a good way to look for typographical errors in your programs. You would also use this option if you have a large program with a lot of functions, and you want to be sure that your functions don't inadvertently use global variables that you meant to be local. (This is a particularly easy mistake to make with simple variable names like i, j, etc.)

-W exec file
--exec file
Similar to -f, reads awk program text from file. There are two differences. The fist is that this option also terminates option processing; anything else on the command line is passed on directly to the awk program. The second is that command line variable assignments of the form ‘var=value’ are disallowed.

This option is particularly necessary for World Wide Web CGI applications that pass arguments through the URL; using this option prevents a malicious (or other) user from passing in options, assignments, or awk source code (via --source) to the CGI application. This option should be used with ‘#!’ scripts (see Executable Scripts), like so:

          #! /usr/local/bin/gawk --exec
          awk program here ...

-W gen-po
Analyzes the source program and generates a GNU gettext Portable Object file on standard output for all string constants that have been marked for translation. See Internationalization, for information about this option.
-W help
-W usage
Prints a “usage” message summarizing the short and long style options that gawk accepts and then exit.
-W lint[=fatal]
Warns about constructs that are dubious or nonportable to other awk implementations. Some warnings are issued when gawk first reads your program. Others are issued at runtime, as your program executes. With an optional argument of ‘fatal’, lint warnings become fatal errors. This may be drastic, but its use will certainly encourage the development of cleaner awk programs. With an optional argument of ‘invalid’, only warnings about things that are actually invalid are issued. (This is not fully implemented yet.)

Some warnings are only printed once, even if the dubious constructs they warn about occur multiple times in your awk program. Thus, when eliminating problems pointed out by --lint, you should take care to search for all occurrences of each inappropriate construct. As awk programs are usually short, doing so is not burdensome.

-W lint-old
Warns about constructs that are not available in the original version of awk from Version 7 Unix (see V7/SVR3.1).
-W non-decimal-data
Enable automatic interpretation of octal and hexadecimal values in input data (see Nondecimal Data).

Caution: This option can severely break old programs. Use with care.

-W posix
Operates in strict POSIX mode. This disables all gawk extensions (just like --traditional) and adds the following additional restrictions:

If you supply both --traditional and --posix on the command line, --posix takes precedence. gawk also issues a warning if both options are supplied.

-W profile[=file]
Enable profiling of awk programs (see Profiling). By default, profiles are created in a file named awkprof.out. The optional file argument allows you to specify a different file name for the profile file.

When run with gawk, the profile is just a “pretty printed” version of the program. When run with pgawk, the profile contains execution counts for each statement in the program in the left margin, and function call counts for each function.

-W re-interval
Allows interval expressions (see Regexp Operators) in regexps. Because interval expressions were traditionally not available in awk, gawk does not provide them by default. This prevents old awk programs from breaking.
-W source program-text
--source program-text
Allows you to mix source code in files with source code that you enter on the command line. Program source code is taken from the program-text. This is particularly useful when you have library functions that you want to use from your command-line programs (see AWKPATH Variable).
-W use-lc-numeric
This option forces the use of the locale's decimal point character when parsing numeric input data (see Locales).
-W version
Prints version information for this particular copy of gawk. This allows you to determine if your copy of gawk is up to date with respect to whatever the Free Software Foundation is currently distributing. It is also useful for bug reports (see Bugs).

As long as program text has been supplied, any other options are flagged as invalid with a warning message but are otherwise ignored.

In compatibility mode, as a special case, if the value of fs supplied to the -F option is ‘t’, then FS is set to the TAB character ("\t"). This is true only for --traditional and not for --posix (see Field Separators).

The -f option may be used more than once on the command line. If it is, awk reads its program source from all of the named files, as if they had been concatenated together into one big file. This is useful for creating libraries of awk functions. These functions can be written once and then retrieved from a standard place, instead of having to be included into each individual program. (As mentioned in Definition Syntax, function names must be unique.)

Library functions can still be used, even if the program is entered at the terminal, by specifying ‘-f /dev/tty’. After typing your program, type Ctrl-d (the end-of-file character) to terminate it. (You may also use ‘-f -’ to read program source from the standard input but then you will not be able to also use the standard input as a source of data.)

Because it is clumsy using the standard awk mechanisms to mix source file and command-line awk programs, gawk provides the --source option. This does not require you to pre-empt the standard input for your source code; it allows you to easily mix command-line and library source code (see AWKPATH Variable).

If no -f or --source option is specified, then gawk uses the first non-option command-line argument as the text of the program source code.

If the environment variable POSIXLY_CORRECT exists, then gawk behaves in strict POSIX mode, exactly as if you had supplied the --posix command-line option. Many GNU programs look for this environment variable to turn on strict POSIX mode. If --lint is supplied on the command line and gawk turns on POSIX mode because of POSIXLY_CORRECT, then it issues a warning message indicating that POSIX mode is in effect. You would typically set this variable in your shell's startup file. For a Bourne-compatible shell (such as bash), you would add these lines to the .profile file in your home directory:


For a csh-compatible shell,53 you would add this line to the .login file in your home directory:

     setenv POSIXLY_CORRECT true

Having POSIXLY_CORRECT set is not recommended for daily use, but it is good for testing the portability of your programs to other environments.

Next: , Previous: Options, Up: Invoking Gawk

11.3 Other Command-Line Arguments

Any additional arguments on the command line are normally treated as input files to be processed in the order specified. However, an argument that has the form var=value, assigns the value value to the variable var—it does not specify a file at all. (This was discussed earlier in Assignment Options.)

All these arguments are made available to your awk program in the ARGV array (see Built-in Variables). Command-line options and the program text (if present) are omitted from ARGV. All other arguments, including variable assignments, are included. As each element of ARGV is processed, gawk sets the variable ARGIND to the index in ARGV of the current element.

The distinction between file name arguments and variable-assignment arguments is made when awk is about to open the next input file. At that point in execution, it checks the file name to see whether it is really a variable assignment; if so, awk sets the variable instead of reading a file.

Therefore, the variables actually receive the given values after all previously specified files have been read. In particular, the values of variables assigned in this fashion are not available inside a BEGIN rule (see BEGIN/END), because such rules are run before awk begins scanning the argument list.

The variable values given on the command line are processed for escape sequences (see Escape Sequences). (d.c.)

In some earlier implementations of awk, when a variable assignment occurred before any file names, the assignment would happen before the BEGIN rule was executed. awk's behavior was thus inconsistent; some command-line assignments were available inside the BEGIN rule, while others were not. Unfortunately, some applications came to depend upon this “feature.” When awk was changed to be more consistent, the -v option was added to accommodate applications that depended upon the old behavior.

The variable assignment feature is most useful for assigning to variables such as RS, OFS, and ORS, which control input and output formats before scanning the data files. It is also useful for controlling state if multiple passes are needed over a data file. For example:

     awk 'pass == 1  { pass 1 stuff }
          pass == 2  { pass 2 stuff }' pass=1 mydata pass=2 mydata

Given the variable assignment feature, the -F option for setting the value of FS is not strictly necessary. It remains for historical compatibility.

Next: , Previous: Other Arguments, Up: Invoking Gawk

11.4 The AWKPATH Environment Variable

In most awk implementations, you must supply a precise path name for each program file, unless the file is in the current directory. But in gawk, if the file name supplied to the -f option does not contain a ‘/’, then gawk searches a list of directories (called the search path), one by one, looking for a file with the specified name.

The search path is a string consisting of directory names separated by colons. gawk gets its search path from the AWKPATH environment variable. If that variable does not exist, gawk uses a default path, ‘.:/usr/local/share/awk’.54 (Programs written for use by system administrators should use an AWKPATH variable that does not include the current directory, ..)

The search path feature is particularly useful for building libraries of useful awk functions. The library files can be placed in a standard directory in the default path and then specified on the command line with a short file name. Otherwise, the full file name would have to be typed for each file.

By using both the --source and -f options, your command-line awk programs can use facilities in awk library files (see Library Functions). Path searching is not done if gawk is in compatibility mode. This is true for both --traditional and --posix. See Options.

NOTE: If you want files in the current directory to be found, you must include the current directory in the path, either by including . explicitly in the path or by writing a null entry in the path. (A null entry is indicated by starting or ending the path with a colon or by placing two colons next to each other (‘::’).) If the current directory is not included in the path, then files cannot be found in the current directory. This path search mechanism is identical to the shell's.

Starting with version 3.0, if AWKPATH is not defined in the environment, gawk places its default search path into ENVIRON["AWKPATH"]. This makes it easy to determine the actual search path that gawk will use from within an awk program.

While you can change ENVIRON["AWKPATH"] within your awk program, this has no effect on the running program's behavior. This makes sense: the AWKPATH environment variable is used to find the program source files. Once your program is running, all the files have been found, and gawk no longer needs to use AWKPATH.

Next: , Previous: AWKPATH Variable, Up: Invoking Gawk

11.5 Obsolete Options and/or Features

This section describes features and/or command-line options from previous releases of gawk that are either not available in the current version or that are still supported but deprecated (meaning that they will not be in the next release).

For version 3.1 of gawk, there are no deprecated command-line options from the previous version of gawk. The use of ‘next file’ (two words) for nextfile was deprecated in gawk 3.0 but still worked. Starting with version 3.1, the two-word usage is no longer accepted.

The process-related special files described in Special Process, work as described, but are now considered deprecated. gawk prints a warning message every time they are used. (Use PROCINFO instead; see Auto-set.) They will be removed from the next release of gawk.

Next: , Previous: Obsolete, Up: Invoking Gawk

11.6 Undocumented Options and Features

Use the Source, Luke!

This section intentionally left blank.

Previous: Undocumented, Up: Invoking Gawk

11.7 Known Bugs in gawk

Next: , Previous: Invoking Gawk, Up: Top

12 A Library of awk Functions

User-defined, describes how to write your own awk functions. Writing functions is important, because it allows you to encapsulate algorithms and program tasks in a single place. It simplifies programming, making program development more manageable, and making programs more readable.

One valuable way to learn a new programming language is to read programs in that language. To that end, this chapter and Sample Programs, provide a good-sized body of code for you to read, and hopefully, to learn from.

This chapter presents a library of useful awk functions. Many of the sample programs presented later in this Web page use these functions. The functions are presented here in a progression from simple to complex.

Extract Program, presents a program that you can use to extract the source code for these example library functions and programs from the Texinfo source for this Web page. (This has already been done as part of the gawk distribution.)

If you have written one or more useful, general-purpose awk functions and would like to contribute them to the author's collection of awk programs, see How To Contribute, for more information.

The programs in this chapter and in Sample Programs, freely use features that are gawk-specific. Rewriting these programs for different implementations of awk is pretty straightforward.

Diagnostic error messages are sent to /dev/stderr. Use ‘| "cat 1>&2"’ instead of ‘> "/dev/stderr"’ if your system does not have a /dev/stderr, or if you cannot use gawk.

A number of programs use nextfile (see Nextfile Statement) to skip any remaining input in the input file. Nextfile Function, shows you how to write a function that does the same thing.

Finally, some of the programs choose to ignore upper- and lowercase distinctions in their input. They do so by assigning one to IGNORECASE. You can achieve almost the same effect55 by adding the following rule to the beginning of the program:

     # ignore case
     { $0 = tolower($0) }

Also, verify that all regexp and string constants used in comparisons use only lowercase letters.

Next: , Up: Library Functions

12.1 Naming Library Function Global Variables

Due to the way the awk language evolved, variables are either global (usable by the entire program) or local (usable just by a specific function). There is no intermediate state analogous to static variables in C.

Library functions often need to have global variables that they can use to preserve state information between calls to the function—for example, getopt's variable _opti (see Getopt Function). Such variables are called private, since the only functions that need to use them are the ones in the library.

When writing a library function, you should try to choose names for your private variables that will not conflict with any variables used by either another library function or a user's main program. For example, a name like ‘i’ or ‘j’ is not a good choice, because user programs often use variable names like these for their own purposes.

The example programs shown in this chapter all start the names of their private variables with an underscore (‘_’). Users generally don't use leading underscores in their variable names, so this convention immediately decreases the chances that the variable name will be accidentally shared with the user's program.

In addition, several of the library functions use a prefix that helps indicate what function or set of functions use the variables—for example, _pw_byname in the user database routines (see Passwd Functions). This convention is recommended, since it even further decreases the chance of inadvertent conflict among variable names. Note that this convention is used equally well for variable names and for private function names as well.56

As a final note on variable naming, if a function makes global variables available for use by a main program, it is a good convention to start that variable's name with a capital letter—for example, getopt's Opterr and Optind variables (see Getopt Function). The leading capital letter indicates that it is global, while the fact that the variable name is not all capital letters indicates that the variable is not one of awk's built-in variables, such as FS.

It is also important that all variables in library functions that do not need to save state are, in fact, declared local.57 If this is not done, the variable could accidentally be used in the user's program, leading to bugs that are very difficult to track down:

     function lib_func(x, y,    l1, l2)
         use variable some_var   # some_var should be local
         ...                   # but is not by oversight

A different convention, common in the Tcl community, is to use a single associative array to hold the values needed by the library function(s), or “package.” This significantly decreases the number of actual global names in use. For example, the functions described in Passwd Functions, might have used array elements PW_data["inited"], PW_data["total"], PW_data["count"], and PW_data["awklib"], instead of _pw_inited, _pw_awklib, _pw_total, and _pw_count.

The conventions presented in this section are exactly that: conventions. You are not required to write your programs this way—we merely recommend that you do so.

Next: , Previous: Library Names, Up: Library Functions

12.2 General Programming

This section presents a number of functions that are of general programming use.

Next: , Up: General Functions

12.2.1 Implementing nextfile as a Function

The nextfile statement, presented in Nextfile Statement, is a gawk-specific extension—it is not available in most other implementations of awk. This section shows two versions of a nextfile function that you can use to simulate gawk's nextfile statement if you cannot use gawk.

A first attempt at writing a nextfile function is as follows:

     # nextfile --- skip remaining records in current file
     # this should be read in before the "main" awk program
     function nextfile()    { _abandon_ = FILENAME; next }
     _abandon_ == FILENAME  { next }

Because it supplies a rule that must be executed first, this file should be included before the main program. This rule compares the current data file's name (which is always in the FILENAME variable) to a private variable named _abandon_. If the file name matches, then the action part of the rule executes a next statement to go on to the next record. (The use of ‘_’ in the variable name is a convention. It is discussed more fully in Library Names.)

The use of the next statement effectively creates a loop that reads all the records from the current data file. The end of the file is eventually reached and a new data file is opened, changing the value of FILENAME. Once this happens, the comparison of _abandon_ to FILENAME fails, and execution continues with the first rule of the “real” program.

The nextfile function itself simply sets the value of _abandon_ and then executes a next statement to start the loop.

This initial version has a subtle problem. If the same data file is listed twice on the command line, one right after the other or even with just a variable assignment between them, this code skips right through the file a second time, even though it should stop when it gets to the end of the first occurrence. A second version of nextfile that remedies this problem is shown here:

     # nextfile --- skip remaining records in current file
     # correctly handle successive occurrences of the same file
     # this should be read in before the "main" awk program
     function nextfile()   { _abandon_ = FILENAME; next }
     _abandon_ == FILENAME {
           if (FNR == 1)
               _abandon_ = ""

The nextfile function has not changed. It makes _abandon_ equal to the current file name and then executes a next statement. The next statement reads the next record and increments FNR so that FNR is guaranteed to have a value of at least two. However, if nextfile is called for the last record in the file, then awk closes the current data file and moves on to the next one. Upon doing so, FILENAME is set to the name of the new file and FNR is reset to one. If this next file is the same as the previous one, _abandon_ is still equal to FILENAME. However, FNR is equal to one, telling us that this is a new occurrence of the file and not the one we were reading when the nextfile function was executed. In that case, _abandon_ is reset to the empty string, so that further executions of this rule fail (until the next time that nextfile is called).

If FNR is not one, then we are still in the original data file and the program executes a next statement to skip through it.

An important question to ask at this point is: given that the functionality of nextfile can be provided with a library file, why is it built into gawk? Adding features for little reason leads to larger, slower programs that are harder to maintain. The answer is that building nextfile into gawk provides significant gains in efficiency. If the nextfile function is executed at the beginning of a large data file, awk still has to scan the entire file, splitting it up into records, just to skip over it. The built-in nextfile can simply close the file immediately and proceed to the next one, which saves a lot of time. This is particularly important in awk, because awk programs are generally I/O-bound (i.e., they spend most of their time doing input and output, instead of performing computations).

Next: , Previous: Nextfile Function, Up: General Functions

12.2.2 Converting Strings To Numbers

The strtonum function (see String Functions) is a gawk extension. The following function provides an implementation for other versions of awk:

     # strtonum --- convert string to number
     function mystrtonum(str,        ret, chars, n, i, k, c)
         if (str ~ /^0[0-7]*$/) {
             # octal
             n = length(str)
             ret = 0
             for (i = 1; i <= n; i++) {
                 c = substr(str, i, 1)
                 if ((k = index("01234567", c)) > 0)
                     k-- # adjust for 1-basing in awk
                 ret = ret * 8 + k
         } else if (str ~ /^0[xX][0-9a-fA-f]+/) {
             # hexadecimal
             str = substr(str, 3)    # lop off leading 0x
             n = length(str)
             ret = 0
             for (i = 1; i <= n; i++) {
                 c = substr(str, i, 1)
                 c = tolower(c)
                 if ((k = index("0123456789", c)) > 0)
                     k-- # adjust for 1-basing in awk
                 else if ((k = index("abcdef", c)) > 0)
                     k += 9
                 ret = ret * 16 + k
         } else if (str ~ /^[-+]?([0-9]+([.][0-9]*([Ee][0-9]+)?)?|([.][0-9]+([Ee][-+]?[0-9]+)?))$/) {
             # decimal number, possibly floating point
             ret = str + 0
         } else
             ret = "NOT-A-NUMBER"
         return ret
     # BEGIN {     # gawk test harness
     #     a[1] = "25"
     #     a[2] = ".31"
     #     a[3] = "0123"
     #     a[4] = "0xdeadBEEF"
     #     a[5] = "123.45"
     #     a[6] = "1.e3"
     #     a[7] = "1.32"
     #     a[7] = "1.32E2"
     #     for (i = 1; i in a; i++)
     #         print a[i], strtonum(a[i]), mystrtonum(a[i])
     # }

The function first looks for C-style octal numbers (base 8). If the input string matches a regular expression describing octal numbers, then mystrtonum loops through each character in the string. It sets k to the index in "01234567" of the current octal digit. Since the return value is one-based, the ‘k--’ adjusts k so it can be used in computing the return value.

Similar logic applies to the code that checks for and converts a hexadecimal value, which starts with ‘0x’ or ‘0X’. The use of tolower simplifies the computation for finding the correct numeric value for each hexadecimal digit.

Finally, if the string matches the (rather complicated) regex for a regular decimal integer or floating-point number, the computation ‘ret = str + 0’ lets awk convert the value to a number.

A commented-out test program is included, so that the function can be tested with gawk and the results compared to the built-in strtonum function.

Next: , Previous: Strtonum Function, Up: General Functions

12.2.3 Assertions

When writing large programs, it is often useful to know that a condition or set of conditions is true. Before proceeding with a particular computation, you make a statement about what you believe to be the case. Such a statement is known as an assertion. The C language provides an <assert.h> header file and corresponding assert macro that the programmer can use to make assertions. If an assertion fails, the assert macro arranges to print a diagnostic message describing the condition that should have been true but was not, and then it kills the program. In C, using assert looks this:

     #include <assert.h>
     int myfunc(int a, double b)
          assert(a <= 5 && b >= 17.1);

If the assertion fails, the program prints a message similar to this:

     prog.c:5: assertion failed: a <= 5 && b >= 17.1

The C language makes it possible to turn the condition into a string for use in printing the diagnostic message. This is not possible in awk, so this assert function also requires a string version of the condition that is being tested. Following is the function:

     # assert --- assert that a condition is true. Otherwise exit.
     function assert(condition, string)
         if (! condition) {
             printf("%s:%d: assertion failed: %s\n",
                 FILENAME, FNR, string) > "/dev/stderr"
             _assert_exit = 1
             exit 1
     END {
         if (_assert_exit)
             exit 1

The assert function tests the condition parameter. If it is false, it prints a message to standard error, using the string parameter to describe the failed condition. It then sets the variable _assert_exit to one and executes the exit statement. The exit statement jumps to the END rule. If the END rules finds _assert_exit to be true, it exits immediately.

The purpose of the test in the END rule is to keep any other END rules from running. When an assertion fails, the program should exit immediately. If no assertions fail, then _assert_exit is still false when the END rule is run normally, and the rest of the program's END rules execute. For all of this to work correctly, assert.awk must be the first source file read by awk. The function can be used in a program in the following way:

     function myfunc(a, b)
          assert(a <= 5 && b >= 17.1, "a <= 5 && b >= 17.1")

If the assertion fails, you see a message similar to the following:

     mydata:1357: assertion failed: a <= 5 && b >= 17.1

There is a small problem with this version of assert. An END rule is automatically added to the program calling assert. Normally, if a program consists of just a BEGIN rule, the input files and/or standard input are not read. However, now that the program has an END rule, awk attempts to read the input data files or standard input (see Using BEGIN/END), most likely causing the program to hang as it waits for input.

There is a simple workaround to this: make sure the BEGIN rule always ends with an exit statement.

Next: , Previous: Assert Function, Up: General Functions

12.2.4 Rounding Numbers

The way printf and sprintf (see Printf) perform rounding often depends upon the system's C sprintf subroutine. On many machines, sprintf rounding is “unbiased,” which means it doesn't always round a trailing ‘.5’ up, contrary to naive expectations. In unbiased rounding, ‘.5’ rounds to even, rather than always up, so 1.5 rounds to 2 but 4.5 rounds to 4. This means that if you are using a format that does rounding (e.g., "%.0f"), you should check what your system does. The following function does traditional rounding; it might be useful if your awk's printf does unbiased rounding:

     # round.awk --- do normal rounding
     function round(x,   ival, aval, fraction)
        ival = int(x)    # integer part, int() truncates
        # see if fractional part
        if (ival == x)   # no fraction
           return x
        if (x < 0) {
           aval = -x     # absolute value
           ival = int(aval)
           fraction = aval - ival
           if (fraction >= .5)
              return int(x) - 1   # -2.5 --> -3
              return int(x)       # -2.3 --> -2
        } else {
           fraction = x - ival
           if (fraction >= .5)
              return ival + 1
              return ival
     # test harness
     { print $0, round($0) }

Next: , Previous: Round Function, Up: General Functions

12.2.5 The Cliff Random Number Generator

The Cliff random number generator58 is a very simple random number generator that “passes the noise sphere test for randomness by showing no structure.” It is easily programmed, in less than 10 lines of awk code:

     # cliff_rand.awk --- generate Cliff random numbers
     BEGIN { _cliff_seed = 0.1 }
     function cliff_rand()
         _cliff_seed = (100 * log(_cliff_seed)) % 1
         if (_cliff_seed < 0)
             _cliff_seed = - _cliff_seed
         return _cliff_seed

This algorithm requires an initial “seed” of 0.1. Each new value uses the current seed as input for the calculation. If the built-in rand function (see Numeric Functions) isn't random enough, you might try using this function instead.

Next: , Previous: Cliff Random Function, Up: General Functions

12.2.6 Translating Between Characters and Numbers

One commercial implementation of awk supplies a built-in function, ord, which takes a character and returns the numeric value for that character in the machine's character set. If the string passed to ord has more than one character, only the first one is used.

The inverse of this function is chr (from the function of the same name in Pascal), which takes a number and returns the corresponding character. Both functions are written very nicely in awk; there is no real reason to build them into the awk interpreter:

     # ord.awk --- do ord and chr
     # Global identifiers:
     #    _ord_:        numerical values indexed by characters
     #    _ord_init:    function to initialize _ord_
     BEGIN    { _ord_init() }
     function _ord_init(    low, high, i, t)
         low = sprintf("%c", 7) # BEL is ascii 7
         if (low == "\a") {    # regular ascii
             low = 0
             high = 127
         } else if (sprintf("%c", 128 + 7) == "\a") {
             # ascii, mark parity
             low = 128
             high = 255
         } else {        # ebcdic(!)
             low = 0
             high = 255
         for (i = low; i <= high; i++) {
             t = sprintf("%c", i)
             _ord_[t] = i

Some explanation of the numbers used by chr is worthwhile. The most prominent character set in use today is ASCII. Although an 8-bit byte can hold 256 distinct values (from 0 to 255), ASCII only defines characters that use the values from 0 to 127.59 In the now distant past, at least one minicomputer manufacturer used ASCII, but with mark parity, meaning that the leftmost bit in the byte is always 1. This means that on those systems, characters have numeric values from 128 to 255. Finally, large mainframe systems use the EBCDIC character set, which uses all 256 values. While there are other character sets in use on some older systems, they are not really worth worrying about:

     function ord(str,    c)
         # only first character is of interest
         c = substr(str, 1, 1)
         return _ord_[c]
     function chr(c)
         # force c to be numeric by adding 0
         return sprintf("%c", c + 0)
     #### test code ####
     # BEGIN    \
     # {
     #    for (;;) {
     #        printf("enter a character: ")
     #        if (getline var <= 0)
     #            break
     #        printf("ord(%s) = %d\n", var, ord(var))
     #    }
     # }

An obvious improvement to these functions is to move the code for the _ord_init function into the body of the BEGIN rule. It was written this way initially for ease of development. There is a “test program” in a BEGIN rule, to test the function. It is commented out for production use.

Next: , Previous: Ordinal Functions, Up: General Functions

12.2.7 Merging an Array into a String

When doing string processing, it is often useful to be able to join all the strings in an array into one long string. The following function, join, accomplishes this task. It is used later in several of the application programs (see Sample Programs).

Good function design is important; this function needs to be general but it should also have a reasonable default behavior. It is called with an array as well as the beginning and ending indices of the elements in the array to be merged. This assumes that the array indices are numeric—a reasonable assumption since the array was likely created with split (see String Functions):

     # join.awk --- join an array into a string
     function join(array, start, end, sep,    result, i)
         if (sep == "")
            sep = " "
         else if (sep == SUBSEP) # magic value
            sep = ""
         result = array[start]
         for (i = start + 1; i <= end; i++)
             result = result sep array[i]
         return result

An optional additional argument is the separator to use when joining the strings back together. If the caller supplies a nonempty value, join uses it; if it is not supplied, it has a null value. In this case, join uses a single blank as a default separator for the strings. If the value is equal to SUBSEP, then join joins the strings with no separator between them. SUBSEP serves as a “magic” value to indicate that there should be no separation between the component strings.60

Previous: Join Function, Up: General Functions

12.2.8 Managing the Time of Day

The systime and strftime functions described in Time Functions, provide the minimum functionality necessary for dealing with the time of day in human readable form. While strftime is extensive, the control formats are not necessarily easy to remember or intuitively obvious when reading a program.

The following function, gettimeofday, populates a user-supplied array with preformatted time information. It returns a string with the current time formatted in the same way as the date utility:

     # gettimeofday.awk --- get the time of day in a usable format
     # Returns a string in the format of output of date(1)
     # Populates the array argument time with individual values:
     #    time["second"]       -- seconds (0 - 59)
     #    time["minute"]       -- minutes (0 - 59)
     #    time["hour"]         -- hours (0 - 23)
     #    time["althour"]      -- hours (0 - 12)
     #    time["monthday"]     -- day of month (1 - 31)
     #    time["month"]        -- month of year (1 - 12)
     #    time["monthname"]    -- name of the month
     #    time["shortmonth"]   -- short name of the month
     #    time["year"]         -- year modulo 100 (0 - 99)
     #    time["fullyear"]     -- full year
     #    time["weekday"]      -- day of week (Sunday = 0)
     #    time["altweekday"]   -- day of week (Monday = 0)
     #    time["dayname"]      -- name of weekday
     #    time["shortdayname"] -- short name of weekday
     #    time["yearday"]      -- day of year (0 - 365)
     #    time["timezone"]     -- abbreviation of timezone name
     #    time["ampm"]         -- AM or PM designation
     #    time["weeknum"]      -- week number, Sunday first day
     #    time["altweeknum"]   -- week number, Monday first day
     function gettimeofday(time,    ret, now, i)
         # get time once, avoids unnecessary system calls
         now = systime()
         # return date(1)-style output
         ret = strftime("%a %b %d %H:%M:%S %Z %Y", now)
         # clear out target array
         delete time
         # fill in values, force numeric values to be
         # numeric by adding 0
         time["second"]       = strftime("%S", now) + 0
         time["minute"]       = strftime("%M", now) + 0
         time["hour"]         = strftime("%H", now) + 0
         time["althour"]      = strftime("%I", now) + 0
         time["monthday"]     = strftime("%d", now) + 0
         time["month"]        = strftime("%m", now) + 0
         time["monthname"]    = strftime("%B", now)
         time["shortmonth"]   = strftime("%b", now)
         time["year"]         = strftime("%y", now) + 0
         time["fullyear"]     = strftime("%Y", now) + 0
         time["weekday"]      = strftime("%w", now) + 0
         time["altweekday"]   = strftime("%u", now) + 0
         time["dayname"]      = strftime("%A", now)
         time["shortdayname"] = strftime("%a", now)
         time["yearday"]      = strftime("%j", now) + 0
         time["timezone"]     = strftime("%Z", now)
         time["ampm"]         = strftime("%p", now)
         time["weeknum"]      = strftime("%U", now) + 0
         time["altweeknum"]   = strftime("%W", now) + 0
         return ret

The string indices are easier to use and read than the various formats required by strftime. The alarm program presented in Alarm Program, uses this function. A more general design for the gettimeofday function would have allowed the user to supply an optional timestamp value to use instead of the current time.

Next: , Previous: General Functions, Up: Library Functions

12.3 Data File Management

This section presents functions that are useful for managing command-line data files.

Next: , Up: Data File Management

12.3.1 Noting Data File Boundaries

The BEGIN and END rules are each executed exactly once at the beginning and end of your awk program, respectively (see BEGIN/END). We (the gawk authors) once had a user who mistakenly thought that the BEGIN rule is executed at the beginning of each data file and the END rule is executed at the end of each data file. When informed that this was not the case, the user requested that we add new special patterns to gawk, named BEGIN_FILE and END_FILE, that would have the desired behavior. He even supplied us the code to do so.

Adding these special patterns to gawk wasn't necessary; the job can be done cleanly in awk itself, as illustrated by the following library program. It arranges to call two user-supplied functions, beginfile and endfile, at the beginning and end of each data file. Besides solving the problem in only nine(!) lines of code, it does so portably; this works with any implementation of awk:

     # transfile.awk
     # Give the user a hook for filename transitions
     # The user must supply functions beginfile() and endfile()
     # that each take the name of the file being started or
     # finished, respectively.
     FILENAME != _oldfilename \
         if (_oldfilename != "")
         _oldfilename = FILENAME
     END   { endfile(FILENAME) }

This file must be loaded before the user's “main” program, so that the rule it supplies is executed first.

This rule relies on awk's FILENAME variable that automatically changes for each new data file. The current file name is saved in a private variable, _oldfilename. If FILENAME does not equal _oldfilename, then a new data file is being processed and it is necessary to call endfile for the old file. Because endfile should only be called if a file has been processed, the program first checks to make sure that _oldfilename is not the null string. The program then assigns the current file name to _oldfilename and calls beginfile for the file. Because, like all awk variables, _oldfilename is initialized to the null string, this rule executes correctly even for the first data file.

The program also supplies an END rule to do the final processing for the last file. Because this END rule comes before any END rules supplied in the “main” program, endfile is called first. Once again the value of multiple BEGIN and END rules should be clear.

This version has same problem as the first version of nextfile (see Nextfile Function). If the same data file occurs twice in a row on the command line, then endfile and beginfile are not executed at the end of the first pass and at the beginning of the second pass. The following version solves the problem:

     # ftrans.awk --- handle data file transitions
     # user supplies beginfile() and endfile() functions
     FNR == 1 {
         if (_filename_ != "")
         _filename_ = FILENAME
     END  { endfile(_filename_) }

Wc Program, shows how this library function can be used and how it simplifies writing the main program.

Next: , Previous: Filetrans Function, Up: Data File Management

12.3.2 Rereading the Current File

Another request for a new built-in function was for a rewind function that would make it possible to reread the current file. The requesting user didn't want to have to use getline (see Getline) inside a loop.

However, as long as you are not in the END rule, it is quite easy to arrange to immediately close the current input file and then start over with it from the top. For lack of a better name, we'll call it rewind:

     # rewind.awk --- rewind the current file and start over
     function rewind(    i)
         # shift remaining arguments up
         for (i = ARGC; i > ARGIND; i--)
             ARGV[i] = ARGV[i-1]
         # make sure gawk knows to keep going
         # make current file next to get done
         # do it

This code relies on the ARGIND variable (see Auto-set), which is specific to gawk. If you are not using gawk, you can use ideas presented in the previous section to either update ARGIND on your own or modify this code as appropriate.

The rewind function also relies on the nextfile keyword (see Nextfile Statement). See Nextfile Function, for a function version of nextfile.

Next: , Previous: Rewind Function, Up: Data File Management

12.3.3 Checking for Readable Data Files

Normally, if you give awk a data file that isn't readable, it stops with a fatal error. There are times when you might want to just ignore such files and keep going. You can do this by prepending the following program to your awk program:

     # readable.awk --- library file to skip over unreadable files
     BEGIN {
         for (i = 1; i < ARGC; i++) {
             if (ARGV[i] ~ /^[A-Za-z_][A-Za-z0-9_]*=.*/ \
                 || ARGV[i] == "-")
                 continue    # assignment or standard input
             else if ((getline junk < ARGV[i]) < 0) # unreadable
                 delete ARGV[i]

In gawk, the getline won't be fatal (unless --posix is in force). Removing the element from ARGV with delete skips the file (since it's no longer in the list).

Next: , Previous: File Checking, Up: Data File Management

12.3.4 Checking For Zero-length Files

All known awk implementations silently skip over zero-length files. This is a by-product of awk's implicit read-a-record-and-match-against-the-rules loop: when awk tries to read a record from an empty file, it immediately receives an end of file indication, closes the file, and proceeds on to the next command-line data file, without executing any user-level awk program code.

Using gawk's ARGIND variable (see Built-in Variables), it is possible to detect when an empty data file has been skipped. Similar to the library file presented in Filetrans Function, the following library file calls a function named zerofile that the user must provide. The arguments passed are the file name and the position in ARGV where it was found:

     # zerofile.awk --- library file to process empty input files
     BEGIN { Argind = 0 }
     ARGIND > Argind + 1 {
         for (Argind++; Argind < ARGIND; Argind++)
             zerofile(ARGV[Argind], Argind)
     ARGIND != Argind { Argind = ARGIND }
     END {
         if (ARGIND > Argind)
             for (Argind++; Argind <= ARGIND; Argind++)
                 zerofile(ARGV[Argind], Argind)

The user-level variable Argind allows the awk program to track its progress through ARGV. Whenever the program detects that ARGIND is greater than ‘Argind + 1’, it means that one or more empty files were skipped. The action then calls zerofile for each such file, incrementing Argind along the way.

The ‘Argind != ARGIND’ rule simply keeps Argind up to date in the normal case.

Finally, the END rule catches the case of any empty files at the end of the command-line arguments. Note that the test in the condition of the for loop uses the ‘<=’ operator, not <.

As an exercise, you might consider whether this same problem can be solved without relying on gawk's ARGIND variable.

As a second exercise, revise this code to handle the case where an intervening value in ARGV is a variable assignment.

Previous: Empty Files, Up: Data File Management

12.3.5 Treating Assignments as File Names

Occasionally, you might not want awk to process command-line variable assignments (see Assignment Options). In particular, if you have file names that contain an ‘=’ character, awk treats the file name as an assignment, and does not process it.

Some users have suggested an additional command-line option for gawk to disable command-line assignments. However, some simple programming with a library file does the trick:

     # noassign.awk --- library file to avoid the need for a
     # special option that disables command-line assignments
     function disable_assigns(argc, argv,    i)
         for (i = 1; i < argc; i++)
             if (argv[i] ~ /^[A-Za-z_][A-Za-z_0-9]*=.*/)
                 argv[i] = ("./" argv[i])
     BEGIN {
         if (No_command_assign)
             disable_assigns(ARGC, ARGV)

You then run your program this way:

     awk -v No_command_assign=1 -f noassign.awk -f yourprog.awk *

The function works by looping through the arguments. It prepends ‘./’ to any argument that matches the form of a variable assignment, turning that argument into a file name.

The use of No_command_assign allows you to disable command-line assignments at invocation time, by giving the variable a true value. When not set, it is initially zero (i.e., false), so the command-line arguments are left alone.

Next: , Previous: Data File Management, Up: Library Functions

12.4 Processing Command-Line Options

Most utilities on POSIX compatible systems take options, or “switches,” on the command line that can be used to change the way a program behaves. awk is an example of such a program (see Options). Often, options take arguments; i.e., data that the program needs to correctly obey the command-line option. For example, awk's -F option requires a string to use as the field separator. The first occurrence on the command line of either -- or a string that does not begin with ‘-’ ends the options.

Modern Unix systems provide a C function named getopt for processing command-line arguments. The programmer provides a string describing the one-letter options. If an option requires an argument, it is followed in the string with a colon. getopt is also passed the count and values of the command-line arguments and is called in a loop. getopt processes the command-line arguments for option letters. Each time around the loop, it returns a single character representing the next option letter that it finds, or ‘?’ if it finds an invalid option. When it returns −1, there are no options left on the command line.

When using getopt, options that do not take arguments can be grouped together. Furthermore, options that take arguments require that the argument is present. The argument can immediately follow the option letter, or it can be a separate command-line argument.

Given a hypothetical program that takes three command-line options, -a, -b, and -c, where -b requires an argument, all of the following are valid ways of invoking the program:

     prog -a -b foo -c data1 data2 data3
     prog -ac -bfoo -- data1 data2 data3
     prog -acbfoo data1 data2 data3

Notice that when the argument is grouped with its option, the rest of the argument is considered to be the option's argument. In this example, -acbfoo indicates that all of the -a, -b, and -c options were supplied, and that ‘foo’ is the argument to the -b option.

getopt provides four external variables that the programmer can use:

The index in the argument value array (argv) where the first nonoption command-line argument can be found.
The string value of the argument to an option.
Usually getopt prints an error message when it finds an invalid option. Setting opterr to zero disables this feature. (An application might want to print its own error message.)
The letter representing the command-line option.

The following C fragment shows how getopt might process command-line arguments for awk:

     main(int argc, char *argv[])
         /* print our own message */
         opterr = 0;
         while ((c = getopt(argc, argv, "v:f:F:W:")) != -1) {
             switch (c) {
             case 'f':    /* file */
             case 'F':    /* field separator */
             case 'v':    /* variable assignment */
             case 'W':    /* extension */
             case '?':

As a side point, gawk actually uses the GNU getopt_long function to process both normal and GNU-style long options (see Options).

The abstraction provided by getopt is very useful and is quite handy in awk programs as well. Following is an awk version of getopt. This function highlights one of the greatest weaknesses in awk, which is that it is very poor at manipulating single characters. Repeated calls to substr are necessary for accessing individual characters (see String Functions).61

The discussion that follows walks through the code a bit at a time:

     # getopt.awk --- do C library getopt(3) function in awk
     # External variables:
     #    Optind -- index in ARGV of first nonoption argument
     #    Optarg -- string value of argument to current option
     #    Opterr -- if nonzero, print our own diagnostic
     #    Optopt -- current option letter
     # Returns:
     #    -1     at end of options
     #    ?      for unrecognized option
     #    <c>    a character representing the current option
     # Private Data:
     #    _opti  -- index in multi-flag option, e.g., -abc

The function starts out with a list of the global variables it uses, what the return values are, what they mean, and any global variables that are “private” to this library function. Such documentation is essential for any program, and particularly for library functions.

The getopt function first checks that it was indeed called with a string of options (the options parameter). If options has a zero length, getopt immediately returns −1:

     function getopt(argc, argv, options,    thisopt, i)
         if (length(options) == 0)    # no options given
             return -1
         if (argv[Optind] == "--") {  # all done
             _opti = 0
             return -1
         } else if (argv[Optind] !~ /^-[^: \t\n\f\r\v\b]/) {
             _opti = 0
             return -1

The next thing to check for is the end of the options. A -- ends the command-line options, as does any command-line argument that does not begin with a ‘-’. Optind is used to step through the array of command-line arguments; it retains its value across calls to getopt, because it is a global variable.

The regular expression that is used, /^-[^: \t\n\f\r\v\b]/, is perhaps a bit of overkill; it checks for a ‘-’ followed by anything that is not whitespace and not a colon. If the current command-line argument does not match this pattern, it is not an option, and it ends option processing:

         if (_opti == 0)
             _opti = 2
         thisopt = substr(argv[Optind], _opti, 1)
         Optopt = thisopt
         i = index(options, thisopt)
         if (i == 0) {
             if (Opterr)
                 printf("%c -- invalid option\n",
                                       thisopt) > "/dev/stderr"
             if (_opti >= length(argv[Optind])) {
                 _opti = 0
             } else
             return "?"

The _opti variable tracks the position in the current command-line argument (argv[Optind]). If multiple options are grouped together with one ‘-’ (e.g., -abx), it is necessary to return them to the user one at a time.

If _opti is equal to zero, it is set to two, which is the index in the string of the next character to look at (we skip the ‘-’, which is at position one). The variable thisopt holds the character, obtained with substr. It is saved in Optopt for the main program to use.

If thisopt is not in the options string, then it is an invalid option. If Opterr is nonzero, getopt prints an error message on the standard error that is similar to the message from the C version of getopt.

Because the option is invalid, it is necessary to skip it and move on to the next option character. If _opti is greater than or equal to the length of the current command-line argument, it is necessary to move on to the next argument, so Optind is incremented and _opti is reset to zero. Otherwise, Optind is left alone and _opti is merely incremented.

In any case, because the option is invalid, getopt returns ‘?’. The main program can examine Optopt if it needs to know what the invalid option letter actually is. Continuing on:

         if (substr(options, i + 1, 1) == ":") {
             # get option argument
             if (length(substr(argv[Optind], _opti + 1)) > 0)
                 Optarg = substr(argv[Optind], _opti + 1)
                 Optarg = argv[++Optind]
             _opti = 0
         } else
             Optarg = ""

If the option requires an argument, the option letter is followed by a colon in the options string. If there are remaining characters in the current command-line argument (argv[Optind]), then the rest of that string is assigned to Optarg. Otherwise, the next command-line argument is used (‘-xFOO’ versus ‘-x FOO’). In either case, _opti is reset to zero, because there are no more characters left to examine in the current command-line argument. Continuing:

         if (_opti == 0 || _opti >= length(argv[Optind])) {
             _opti = 0
         } else
         return thisopt

Finally, if _opti is either zero or greater than the length of the current command-line argument, it means this element in argv is through being processed, so Optind is incremented to point to the next element in argv. If neither condition is true, then only _opti is incremented, so that the next option letter can be processed on the next call to getopt.

The BEGIN rule initializes both Opterr and Optind to one. Opterr is set to one, since the default behavior is for getopt to print a diagnostic message upon seeing an invalid option. Optind is set to one, since there's no reason to look at the program name, which is in ARGV[0]:

     BEGIN {
         Opterr = 1    # default is to diagnose
         Optind = 1    # skip ARGV[0]
         # test program
         if (_getopt_test) {
             while ((_go_c = getopt(ARGC, ARGV, "ab:cd")) != -1)
                 printf("c = <%c>, optarg = <%s>\n",
                                            _go_c, Optarg)
             printf("non-option arguments:\n")
             for (; Optind < ARGC; Optind++)
                 printf("\tARGV[%d] = <%s>\n",
                                         Optind, ARGV[Optind])

The rest of the BEGIN rule is a simple test program. Here is the result of two sample runs of the test program:

     $ awk -f getopt.awk -v _getopt_test=1 -- -a -cbARG bax -x
     -| c = <a>, optarg = <>
     -| c = <c>, optarg = <>
     -| c = <b>, optarg = <ARG>
     -| non-option arguments:
     -|         ARGV[3] = <bax>
     -|         ARGV[4] = <-x>
     $ awk -f getopt.awk -v _getopt_test=1 -- -a -x -- xyz abc
     -| c = <a>, optarg = <>
     error--> x -- invalid option
     -| c = <?>, optarg = <>
     -| non-option arguments:
     -|         ARGV[4] = <xyz>
     -|         ARGV[5] = <abc>

In both runs, the first -- terminates the arguments to awk, so that it does not try to interpret the -a, etc., as its own options.

NOTE: After getopt is through, it is the responsibility of the user level code to clear out all the elements of ARGV from 1 to Optind, so that awk does not try to process the command-line options as file names.

Several of the sample programs presented in Sample Programs, use getopt to process their arguments.

Next: , Previous: Getopt Function, Up: Library Functions

12.5 Reading the User Database

The PROCINFO array (see Built-in Variables) provides access to the current user's real and effective user and group ID numbers, and if available, the user's supplementary group set. However, because these are numbers, they do not provide very useful information to the average user. There needs to be some way to find the user information associated with the user and group ID numbers. This section presents a suite of functions for retrieving information from the user database. See Group Functions, for a similar suite that retrieves information from the group database.

The POSIX standard does not define the file where user information is kept. Instead, it provides the <pwd.h> header file and several C language subroutines for obtaining user information. The primary function is getpwent, for “get password entry.” The “password” comes from the original user database file, /etc/passwd, which stores user information, along with the encrypted passwords (hence the name).

While an awk program could simply read /etc/passwd directly, this file may not contain complete information about the system's set of users.62 To be sure you are able to produce a readable and complete version of the user database, it is necessary to write a small C program that calls getpwent. getpwent is defined as returning a pointer to a struct passwd. Each time it is called, it returns the next entry in the database. When there are no more entries, it returns NULL, the null pointer. When this happens, the C program should call endpwent to close the database. Following is pwcat, a C program that “cats” the password database:

      * pwcat.c
      * Generate a printable version of the password database
     #include <stdio.h>
     #include <pwd.h>
     main(argc, argv)
     int argc;
     char **argv;
         struct passwd *p;
         while ((p = getpwent()) != NULL)
                 p->pw_name, p->pw_passwd, (long) p->pw_uid,
                 (long) p->pw_gid, p->pw_gecos, p->pw_dir, p->pw_shell);
         return 0;

If you don't understand C, don't worry about it. The output from pwcat is the user database, in the traditional /etc/passwd format of colon-separated fields. The fields are:

Login name The user's login name.

Encrypted password The user's encrypted password. This may not be available on some systems.

User-ID The user's numeric user ID number.

Group-ID The user's numeric group ID number.

Full name The user's full name, and perhaps other information associated with the user.

Home directory The user's login (or “home”) directory (familiar to shell programmers as $HOME).

Login shell The program that is run when the user logs in. This is usually a shell, such as bash.

A few lines representative of pwcat's output are as follows:

     $ pwcat
     -| root:3Ov02d5VaUPB6:0:1:Operator:/:/bin/sh
     -| nobody:*:65534:65534::/:
     -| daemon:*:1:1::/:
     -| sys:*:2:2::/:/bin/csh
     -| bin:*:3:3::/bin:
     -| arnold:xyzzy:2076:10:Arnold Robbins:/home/arnold:/bin/sh
     -| miriam:yxaay:112:10:Miriam Robbins:/home/miriam:/bin/sh
     -| andy:abcca2:113:10:Andy Jacobs:/home/andy:/bin/sh

With that introduction, following is a group of functions for getting user information. There are several functions here, corresponding to the C functions of the same names:

     # passwd.awk --- access password file information
     BEGIN {
         # tailor this to suit your system
         _pw_awklib = "/usr/local/libexec/awk/"
     function _pw_init(    oldfs, oldrs, olddol0, pwcat, using_fw)
         if (_pw_inited)
         oldfs = FS
         oldrs = RS
         olddol0 = $0
         using_fw = (PROCINFO["FS"] == "FIELDWIDTHS")
         FS = ":"
         RS = "\n"
         pwcat = _pw_awklib "pwcat"
         while ((pwcat | getline) > 0) {
             _pw_byname[$1] = $0
             _pw_byuid[$3] = $0
             _pw_bycount[++_pw_total] = $0
         _pw_count = 0
         _pw_inited = 1
         FS = oldfs
         if (using_fw)
         RS = oldrs
         $0 = olddol0

The BEGIN rule sets a private variable to the directory where pwcat is stored. Because it is used to help out an awk library routine, we have chosen to put it in /usr/local/libexec/awk; however, you might want it to be in a different directory on your system.

The function _pw_init keeps three copies of the user information in three associative arrays. The arrays are indexed by username (_pw_byname), by user ID number (_pw_byuid), and by order of occurrence (_pw_bycount). The variable _pw_inited is used for efficiency; _pw_init needs only to be called once.

Because this function uses getline to read information from pwcat, it first saves the values of FS, RS, and $0. It notes in the variable using_fw whether field splitting with FIELDWIDTHS is in effect or not. Doing so is necessary, since these functions could be called from anywhere within a user's program, and the user may have his or her own way of splitting records and fields.

The using_fw variable checks PROCINFO["FS"], which is "FIELDWIDTHS" if field splitting is being done with FIELDWIDTHS. This makes it possible to restore the correct field-splitting mechanism later. The test can only be true for gawk. It is false if using FS or on some other awk implementation.

The main part of the function uses a loop to read database lines, split the line into fields, and then store the line into each array as necessary. When the loop is done, _pw_init cleans up by closing the pipeline, setting _pw_inited to one, and restoring FS (and FIELDWIDTHS if necessary), RS, and $0. The use of _pw_count is explained shortly.

The getpwnam function takes a username as a string argument. If that user is in the database, it returns the appropriate line. Otherwise, it returns the null string:

     function getpwnam(name)
         if (name in _pw_byname)
             return _pw_byname[name]
         return ""

Similarly, the getpwuid function takes a user ID number argument. If that user number is in the database, it returns the appropriate line. Otherwise, it returns the null string:

     function getpwuid(uid)
         if (uid in _pw_byuid)
             return _pw_byuid[uid]
         return ""

The getpwent function simply steps through the database, one entry at a time. It uses _pw_count to track its current position in the _pw_bycount array:

     function getpwent()
         if (_pw_count < _pw_total)
             return _pw_bycount[++_pw_count]
         return ""

The endpwent function resets _pw_count to zero, so that subsequent calls to getpwent start over again:

     function endpwent()
         _pw_count = 0

A conscious design decision in this suite was made that each subroutine calls _pw_init to initialize the database arrays. The overhead of running a separate process to generate the user database, and the I/O to scan it, are only incurred if the user's main program actually calls one of these functions. If this library file is loaded along with a user's program, but none of the routines are ever called, then there is no extra runtime overhead. (The alternative is move the body of _pw_init into a BEGIN rule, which always runs pwcat. This simplifies the code but runs an extra process that may never be needed.)

In turn, calling _pw_init is not too expensive, because the _pw_inited variable keeps the program from reading the data more than once. If you are worried about squeezing every last cycle out of your awk program, the check of _pw_inited could be moved out of _pw_init and duplicated in all the other functions. In practice, this is not necessary, since most awk programs are I/O-bound, and it clutters up the code.

The id program in Id Program, uses these functions.

Previous: Passwd Functions, Up: Library Functions

12.6 Reading the Group Database

Much of the discussion presented in Passwd Functions, applies to the group database as well. Although there has traditionally been a well-known file (/etc/group) in a well-known format, the POSIX standard only provides a set of C library routines (<grp.h> and getgrent) for accessing the information. Even though this file may exist, it likely does not have complete information. Therefore, as with the user database, it is necessary to have a small C program that generates the group database as its output.

grcat, a C program that “cats” the group database, is as follows:

      * grcat.c
      * Generate a printable version of the group database
     #include <stdio.h>
     #include <grp.h>
     main(argc, argv)
     int argc;
     char **argv;
         struct group *g;
         int i;
         while ((g = getgrent()) != NULL) {
             printf("%s:%s:%ld:", g->gr_name, g->gr_passwd,
                                          (long) g->gr_gid);
             for (i = 0; g->gr_mem[i] != NULL; i++) {
                 printf("%s", g->gr_mem[i]);
                 if (g->gr_mem[i+1] != NULL)
         return 0;

Each line in the group database represents one group. The fields are separated with colons and represent the following information:

Group name The group's name.

Group password The group's encrypted password. In practice, this field is never used; it is usually empty or set to ‘*’.

Group-ID The group's numeric group ID number; this number should be unique within the file.

Group member list A comma-separated list of user names. These users are members of the group. Modern Unix systems allow users to be members of several groups simultaneously. If your system does, then there are elements "group1" through "groupN" in PROCINFO for those group ID numbers. (Note that PROCINFO is a gawk extension; see Built-in Variables.)

Here is what running grcat might produce:

     $ grcat
     -| wheel:*:0:arnold
     -| nogroup:*:65534:
     -| daemon:*:1:
     -| kmem:*:2:
     -| staff:*:10:arnold,miriam,andy
     -| other:*:20:

Here are the functions for obtaining information from the group database. There are several, modeled after the C library functions of the same names:

     # group.awk --- functions for dealing with the group file
     BEGIN    \
         # Change to suit your system
         _gr_awklib = "/usr/local/libexec/awk/"
     function _gr_init(    oldfs, oldrs, olddol0, grcat,
                                  using_fw, n, a, i)
         if (_gr_inited)
         oldfs = FS
         oldrs = RS
         olddol0 = $0
         using_fw = (PROCINFO["FS"] == "FIELDWIDTHS")
         FS = ":"
         RS = "\n"
         grcat = _gr_awklib "grcat"
         while ((grcat | getline) > 0) {
             if ($1 in _gr_byname)
                 _gr_byname[$1] = _gr_byname[$1] "," $4
                 _gr_byname[$1] = $0
             if ($3 in _gr_bygid)
                 _gr_bygid[$3] = _gr_bygid[$3] "," $4
                 _gr_bygid[$3] = $0
             n = split($4, a, "[ \t]*,[ \t]*")
             for (i = 1; i <= n; i++)
                 if (a[i] in _gr_groupsbyuser)
                     _gr_groupsbyuser[a[i]] = \
                         _gr_groupsbyuser[a[i]] " " $1
                     _gr_groupsbyuser[a[i]] = $1
             _gr_bycount[++_gr_count] = $0
         _gr_count = 0
         FS = oldfs
         if (using_fw)
         RS = oldrs
         $0 = olddol0

The BEGIN rule sets a private variable to the directory where grcat is stored. Because it is used to help out an awk library routine, we have chosen to put it in /usr/local/libexec/awk. You might want it to be in a different directory on your system.

These routines follow the same general outline as the user database routines (see Passwd Functions). The _gr_inited variable is used to ensure that the database is scanned no more than once. The _gr_init function first saves FS, FIELDWIDTHS, RS, and $0, and then sets FS and RS to the correct values for scanning the group information.

The group information is stored is several associative arrays. The arrays are indexed by group name (_gr_byname), by group ID number (_gr_bygid), and by position in the database (_gr_bycount). There is an additional array indexed by user name (_gr_groupsbyuser), which is a space-separated list of groups to which each user belongs.

Unlike the user database, it is possible to have multiple records in the database for the same group. This is common when a group has a large number of members. A pair of such entries might look like the following:


For this reason, _gr_init looks to see if a group name or group ID number is already seen. If it is, then the user names are simply concatenated onto the previous list of users. (There is actually a subtle problem with the code just presented. Suppose that the first time there were no names. This code adds the names with a leading comma. It also doesn't check that there is a $4.)

Finally, _gr_init closes the pipeline to grcat, restores FS (and FIELDWIDTHS if necessary), RS, and $0, initializes _gr_count to zero (it is used later), and makes _gr_inited nonzero.

The getgrnam function takes a group name as its argument, and if that group exists, it is returned. Otherwise, getgrnam returns the null string:

     function getgrnam(group)
         if (group in _gr_byname)
             return _gr_byname[group]
         return ""

The getgrgid function is similar, it takes a numeric group ID and looks up the information associated with that group ID:

     function getgrgid(gid)
         if (gid in _gr_bygid)
             return _gr_bygid[gid]
         return ""

The getgruser function does not have a C counterpart. It takes a user name and returns the list of groups that have the user as a member:

     function getgruser(user)
         if (user in _gr_groupsbyuser)
             return _gr_groupsbyuser[user]
         return ""

The getgrent function steps through the database one entry at a time. It uses _gr_count to track its position in the list:

     function getgrent()
         if (++_gr_count in _gr_bycount)
             return _gr_bycount[_gr_count]
         return ""

The endgrent function resets _gr_count to zero so that getgrent can start over again:

     function endgrent()
         _gr_count = 0

As with the user database routines, each function calls _gr_init to initialize the arrays. Doing so only incurs the extra overhead of running grcat if these functions are used (as opposed to moving the body of _gr_init into a BEGIN rule).

Most of the work is in scanning the database and building the various associative arrays. The functions that the user calls are themselves very simple, relying on awk's associative arrays to do work.

The id program in Id Program, uses these functions.

Next: , Previous: Library Functions, Up: Top

13 Practical awk Programs

Library Functions, presents the idea that reading programs in a language contributes to learning that language. This chapter continues that theme, presenting a potpourri of awk programs for your reading enjoyment. There are three sections. The first describes how to run the programs presented in this chapter.

The second presents awk versions of several common POSIX utilities. These are programs that you are hopefully already familiar with, and therefore, whose problems are understood. By reimplementing these programs in awk, you can focus on the awk-related aspects of solving the programming problem.

The third is a grab bag of interesting programs. These solve a number of different data-manipulation and management problems. Many of the programs are short, which emphasizes awk's ability to do a lot in just a few lines of code.

Many of these programs use the library functions presented in Library Functions.

Next: , Up: Sample Programs

13.1 Running the Example Programs

To run a given program, you would typically do something like this:

     awk -f program -- options files

Here, program is the name of the awk program (such as cut.awk), options are any command-line options for the program that start with a ‘-’, and files are the actual data files.

If your system supports the ‘#!’ executable interpreter mechanism (see Executable Scripts), you can instead run your program directly:

     cut.awk -c1-8 myfiles > results

If your awk is not gawk, you may instead need to use this:

     cut.awk -- -c1-8 myfiles > results

Next: , Previous: Running Examples, Up: Sample Programs

13.2 Reinventing Wheels for Fun and Profit

This section presents a number of POSIX utilities that are implemented in awk. Reinventing these programs in awk is often enjoyable, because the algorithms can be very clearly expressed, and the code is usually very concise and simple. This is true because awk does so much for you.

It should be noted that these programs are not necessarily intended to replace the installed versions on your system. Instead, their purpose is to illustrate awk language programming for “real world” tasks.

The programs are presented in alphabetical order.

Next: , Up: Clones

13.2.1 Cutting out Fields and Columns

The cut utility selects, or “cuts,” characters or fields from its standard input and sends them to its standard output. Fields are separated by tabs by default, but you may supply a command-line option to change the field delimiter (i.e., the field-separator character). cut's definition of fields is less general than awk's.

A common use of cut might be to pull out just the login name of logged-on users from the output of who. For example, the following pipeline generates a sorted, unique list of the logged-on users:

     who | cut -c1-8 | sort | uniq

The options for cut are:

-c list
Use list as the list of characters to cut out. Items within the list may be separated by commas, and ranges of characters can be separated with dashes. The list ‘1-8,15,22-35’ specifies characters 1 through 8, 15, and 22 through 35.
-f list
Use list as the list of fields to cut out.
-d delim
Use delim as the field-separator character instead of the tab character.
Suppress printing of lines that do not contain the field delimiter.

The awk implementation of cut uses the getopt library function (see Getopt Function) and the join library function (see Join Function).

The program begins with a comment describing the options, the library functions needed, and a usage function that prints out a usage message and exits. usage is called if invalid arguments are supplied:

     # cut.awk --- implement cut in awk
     # Options:
     #    -f list     Cut fields
     #    -d c        Field delimiter character
     #    -c list     Cut characters
     #    -s          Suppress lines without the delimiter
     # Requires getopt and join library functions
     function usage(    e1, e2)
         e1 = "usage: cut [-f list] [-d c] [-s] [files...]"
         e2 = "usage: cut [-c list] [files...]"
         print e1 > "/dev/stderr"
         print e2 > "/dev/stderr"
         exit 1

The variables e1 and e2 are used so that the function fits nicely on the page. screen.

Next comes a BEGIN rule that parses the command-line options. It sets FS to a single TAB character, because that is cut's default field separator. The output field separator is also set to be the same as the input field separator. Then getopt is used to step through the command-line options. Exactly one of the variables by_fields or by_chars is set to true, to indicate that processing should be done by fields or by characters, respectively. When cutting by characters, the output field separator is set to the null string:

     BEGIN    \
         FS = "\t"    # default
         OFS = FS
         while ((c = getopt(ARGC, ARGV, "sf:c:d:")) != -1) {
             if (c == "f") {
                 by_fields = 1
                 fieldlist = Optarg
             } else if (c == "c") {
                 by_chars = 1
                 fieldlist = Optarg
                 OFS = ""
             } else if (c == "d") {
                 if (length(Optarg) > 1) {
                     printf("Using first character of %s" \
                     " for delimiter\n", Optarg) > "/dev/stderr"
                     Optarg = substr(Optarg, 1, 1)
                 FS = Optarg
                 OFS = FS
                 if (FS == " ")    # defeat awk semantics
                     FS = "[ ]"
             } else if (c == "s")
         for (i = 1; i < Optind; i++)
             ARGV[i] = ""

Special care is taken when the field delimiter is a space. Using a single space (" ") for the value of FS is incorrect—awk would separate fields with runs of spaces, tabs, and/or newlines, and we want them to be separated with individual spaces. Also remember that after getopt is through (as described in Getopt Function), we have to clear out all the elements of ARGV from 1 to Optind, so that awk does not try to process the command-line options as file names.

After dealing with the command-line options, the program verifies that the options make sense. Only one or the other of -c and -f should be used, and both require a field list. Then the program calls either set_fieldlist or set_charlist to pull apart the list of fields or characters:

         if (by_fields && by_chars)
         if (by_fields == 0 && by_chars == 0)
             by_fields = 1    # default
         if (fieldlist == "") {
             print "cut: needs list for -c or -f" > "/dev/stderr"
             exit 1
         if (by_fields)

set_fieldlist is used to split the field list apart at the commas and into an array. Then, for each element of the array, it looks to see if it is actually a range, and if so, splits it apart. The range is verified to make sure the first number is smaller than the second. Each number in the list is added to the flist array, which simply lists the fields that will be printed. Normal field splitting is used. The program lets awk handle the job of doing the field splitting:

     function set_fieldlist(        n, m, i, j, k, f, g)
         n = split(fieldlist, f, ",")
         j = 1    # index in flist
         for (i = 1; i <= n; i++) {
             if (index(f[i], "-") != 0) { # a range
                 m = split(f[i], g, "-")
                 if (m != 2 || g[1] >= g[2]) {
                     printf("bad field list: %s\n",
                                       f[i]) > "/dev/stderr"
                     exit 1
                 for (k = g[1]; k <= g[2]; k++)
                     flist[j++] = k
             } else
                 flist[j++] = f[i]
         nfields = j - 1

The set_charlist function is more complicated than set_fieldlist. The idea here is to use gawk's FIELDWIDTHS variable (see Constant Size), which describes constant-width input. When using a character list, that is exactly what we have.

Setting up FIELDWIDTHS is more complicated than simply listing the fields that need to be printed. We have to keep track of the fields to print and also the intervening characters that have to be skipped. For example, suppose you wanted characters 1 through 8, 15, and 22 through 35. You would use ‘-c 1-8,15,22-35’. The necessary value for FIELDWIDTHS is "8 6 1 6 14". This yields five fields, and the fields to print are $1, $3, and $5. The intermediate fields are filler, which is stuff in between the desired data. flist lists the fields to print, and t tracks the complete field list, including filler fields:

     function set_charlist(    field, i, j, f, g, t,
                               filler, last, len)
         field = 1   # count total fields
         n = split(fieldlist, f, ",")
         j = 1       # index in flist
         for (i = 1; i <= n; i++) {
             if (index(f[i], "-") != 0) { # range
                 m = split(f[i], g, "-")
                 if (m != 2 || g[1] >= g[2]) {
                     printf("bad character list: %s\n",
                                    f[i]) > "/dev/stderr"
                     exit 1
                 len = g[2] - g[1] + 1
                 if (g[1] > 1)  # compute length of filler
                     filler = g[1] - last - 1
                     filler = 0
                 if (filler)
                     t[field++] = filler
                 t[field++] = len  # length of field
                 last = g[2]
                 flist[j++] = field - 1
             } else {
                 if (f[i] > 1)
                     filler = f[i] - last - 1
                     filler = 0
                 if (filler)
                     t[field++] = filler
                 t[field++] = 1
                 last = f[i]
                 flist[j++] = field - 1
         FIELDWIDTHS = join(t, 1, field - 1)
         nfields = j - 1

Next is the rule that actually processes the data. If the -s option is given, then suppress is true. The first if statement makes sure that the input record does have the field separator. If cut is processing fields, suppress is true, and the field separator character is not in the record, then the record is skipped.

If the record is valid, then gawk has split the data into fields, either using the character in FS or using fixed-length fields and FIELDWIDTHS. The loop goes through the list of fields that should be printed. The corresponding field is printed if it contains data. If the next field also has data, then the separator character is written out between the fields:

         if (by_fields && suppress && index($0, FS) != 0)
         for (i = 1; i <= nfields; i++) {
             if ($flist[i] != "") {
                 printf "%s", $flist[i]
                 if (i < nfields && $flist[i+1] != "")
                     printf "%s", OFS
         print ""

This version of cut relies on gawk's FIELDWIDTHS variable to do the character-based cutting. While it is possible in other awk implementations to use substr (see String Functions), it is also extremely painful. The FIELDWIDTHS variable supplies an elegant solution to the problem of picking the input line apart by characters.

Next: , Previous: Cut Program, Up: Clones

13.2.2 Searching for Regular Expressions in Files

The egrep utility searches files for patterns. It uses regular expressions that are almost identical to those available in awk (see Regexp). It is used in the following manner:

     egrep [ options ] 'pattern' files ...

The pattern is a regular expression. In typical usage, the regular expression is quoted to prevent the shell from expanding any of the special characters as file name wildcards. Normally, egrep prints the lines that matched. If multiple file names are provided on the command line, each output line is preceded by the name of the file and a colon.

The options to egrep are as follows:

Print out a count of the lines that matched the pattern, instead of the lines themselves.
Be silent. No output is produced and the exit value indicates whether the pattern was matched.
Invert the sense of the test. egrep prints the lines that do not match the pattern and exits successfully if the pattern is not matched.
Ignore case distinctions in both the pattern and the input data.
Only print (list) the names of the files that matched, not the lines that matched.
-e pattern
Use pattern as the regexp to match. The purpose of the -e option is to allow patterns that start with a ‘-’.

This version uses the getopt library function (see Getopt Function) and the file transition library program (see Filetrans Function).

The program begins with a descriptive comment and then a BEGIN rule that processes the command-line arguments with getopt. The -i (ignore case) option is particularly easy with gawk; we just use the IGNORECASE built-in variable (see Built-in Variables):

     # egrep.awk --- simulate egrep in awk
     # Options:
     #    -c    count of lines
     #    -s    silent - use exit value
     #    -v    invert test, success if no match
     #    -i    ignore case
     #    -l    print filenames only
     #    -e    argument is pattern
     # Requires getopt and file transition library functions
     BEGIN {
         while ((c = getopt(ARGC, ARGV, "ce:svil")) != -1) {
             if (c == "c")
             else if (c == "s")
             else if (c == "v")
             else if (c == "i")
                 IGNORECASE = 1
             else if (c == "l")
             else if (c == "e")
                 pattern = Optarg

Next comes the code that handles the egrep-specific behavior. If no pattern is supplied with -e, the first nonoption on the command line is used. The awk command-line arguments up to ARGV[Optind] are cleared, so that awk won't try to process them as files. If no files are specified, the standard input is used, and if multiple files are specified, we make sure to note this so that the file names can precede the matched lines in the output:

         if (pattern == "")
             pattern = ARGV[Optind++]
         for (i = 1; i < Optind; i++)
             ARGV[i] = ""
         if (Optind >= ARGC) {
             ARGV[1] = "-"
             ARGC = 2
         } else if (ARGC - Optind > 1)
     #    if (IGNORECASE)
     #        pattern = tolower(pattern)

The last two lines are commented out, since they are not needed in gawk. They should be uncommented if you have to use another version of awk.

The next set of lines should be uncommented if you are not using gawk. This rule translates all the characters in the input line into lowercase if the -i option is specified.63 The rule is commented out since it is not necessary with gawk:

     #    if (IGNORECASE)
     #        $0 = tolower($0)

The beginfile function is called by the rule in ftrans.awk when each new file is processed. In this case, it is very simple; all it does is initialize a variable fcount to zero. fcount tracks how many lines in the current file matched the pattern (naming the parameter junk shows we know that beginfile is called with a parameter, but that we're not interested in its value):

     function beginfile(junk)
         fcount = 0

The endfile function is called after each file has been processed. It affects the output only when the user wants a count of the number of lines that matched. no_print is true only if the exit status is desired. count_only is true if line counts are desired. egrep therefore only prints line counts if printing and counting are enabled. The output format must be adjusted depending upon the number of files to process. Finally, fcount is added to total, so that we know the total number of lines that matched the pattern:

     function endfile(file)
         if (! no_print && count_only)
             if (do_filenames)
                 print file ":" fcount
                 print fcount
         total += fcount

The following rule does most of the work of matching lines. The variable matches is true if the line matched the pattern. If the user wants lines that did not match, the sense of matches is inverted using the ‘!’ operator. fcount is incremented with the value of matches, which is either one or zero, depending upon a successful or unsuccessful match. If the line does not match, the next statement just moves on to the next record.

A number of additional tests are made, but they are only done if we are not counting lines. First, if the user only wants exit status (no_print is true), then it is enough to know that one line in this file matched, and we can skip on to the next file with nextfile. Similarly, if we are only printing file names, we can print the file name, and then skip to the next file with nextfile. Finally, each line is printed, with a leading file name and colon if necessary:

         matches = ($0 ~ pattern)
         if (invert)
             matches = ! matches
         fcount += matches    # 1 or 0
         if (! matches)
         if (! count_only) {
             if (no_print)
             if (filenames_only) {
                 print FILENAME
             if (do_filenames)
                 print FILENAME ":" $0

The END rule takes care of producing the correct exit status. If there are no matches, the exit status is one; otherwise it is zero:

     END    \
         if (total == 0)
             exit 1
         exit 0

The usage function prints a usage message in case of invalid options, and then exits:

     function usage(    e)
         e = "Usage: egrep [-csvil] [-e pat] [files ...]"
         e = e "\n\tegrep [-csvil] pat [files ...]"
         print e > "/dev/stderr"
         exit 1

The variable e is used so that the function fits nicely on the printed page.

Just a note on programming style: you may have noticed that the END rule uses backslash continuation, with the open brace on a line by itself. This is so that it more closely resembles the way functions are written. Many of the examples in this chapter use this style. You can decide for yourself if you like writing your BEGIN and END rules this way or not.

Next: , Previous: Egrep Program, Up: Clones

13.2.3 Printing out User Information

The id utility lists a user's real and effective user ID numbers, real and effective group ID numbers, and the user's group set, if any. id only prints the effective user ID and group ID if they are different from the real ones. If possible, id also supplies the corresponding user and group names. The output might look like this:

     $ id
     -| uid=2076(arnold) gid=10(staff) group