w64devkit/man/man1/grep.1

.TH GREP 1 2026-10-27 "w64devkit" "w64devkit Manual"
.SH NAME
grep, egrep, fgrep \- print lines that match patterns
.SH SYNOPSIS
.B grep
.RI [ options ]
.I pattern
.RI [ file ...]
.br
.B grep
.RI [ options ]
.BI "\-e " pattern
.RI ...\&
.RI [ file ...]
.br
.B grep
.RI [ options ]
.BI "\-f " patfile
.RI ...\&
.RI [ file ...]
.br
.B egrep
.RI [ options ]
.I pattern
.RI [ file ...]
.br
.B fgrep
.RI [ options ]
.I pattern
.RI [ file ...]
.SH DESCRIPTION
.B grep
searches each
.I file
for lines that match any of the patterns, and writes those lines to
standard output.
With no
.I file
operands, or for a
.I file
named
.BR \- ,
it reads standard input.
.B egrep
is
.BR "grep \-E" ,
and
.B fgrep
is
.BR "grep \-F" .
This is the busybox
.BR grep .
.PP
The patterns are those given with
.B \-e
and
.BR \-f ,
or when neither is given, the
.I pattern
operand.
A pattern containing newlines is several patterns, one per line, and an
empty pattern matches every line.
Patterns are basic regular expressions, extended regular expressions
with
.BR \-E ,
or fixed strings with
.BR \-F .
.PP
A line of input ends at an LF or at a NUL byte, and a CR before an LF is
removed, so CRLF text is matched as if it had LF line endings.
Lines may be any length.
Every file is searched as text; there is no binary file detection.
.PP
Each selected line is written followed by an LF.
When more than one
.I file
is given, or a directory is searched with
.BR \-r ,
the line is preceded by its file name and a colon;
.B \-H
and
.B \-h
override this.
Standard input is named
.BR "(standard input)" .
With
.BR \-n ,
the line number and a colon come next.
Context lines, requested with
.BR \-A ,
.BR \-B ,
or
.BR \-C ,
have a
.B \-
in place of each colon, and a line containing only
.B \-\-
separates groups of lines that are not adjacent.
.BR \-c ,
.BR \-l ,
.BR \-L ,
and
.B \-q
replace this output, and they disable
.B \-n
and context.
.SS Regular expressions
Regular expressions are POSIX basic or extended regular expressions, as
implemented by the GNU regex code built into busybox, with its
extensions.
In a basic expression,
.BR \e? ,
.BR \e+ ,
and
.B \e|
are the optional, repetition, and alternation operators that
.BR ? ,
.BR + ,
and
.B |
are in an extended expression, and
.BR ? ,
.BR + ,
and
.B |
match themselves.
An interval is
.BI \e{ m , n \e}
in a basic expression and
.BI { m , n }
in an extended one.
Both kinds accept back-references
.B \e1
to
.BR \e9 ,
.B \ew
for a word character, which is an ASCII letter, digit, or underscore,
.B \eW
for any other character,
.B \es
and
.B \eS
for whitespace and anything else,
.B \eb
and
.B \eB
for a word boundary and its absence, and
.B \e<
and
.B \e>
for the start and end of a word.
.PP
Matching works on bytes.
A
.B .\&
or a bracket expression matches one byte, not one non-ASCII character,
and character classes such as
.B [:alpha:]
contain only ASCII characters.
.SH OPTIONS
Options may be grouped, as in
.BR \-in ,
and may follow operands unless
.B POSIXLY_CORRECT
is set.
An option's argument may be attached, as in
.BR \-m1 ,
or be the next argument.
.B \-\-
ends the options.
The
.I num
arguments are decimal integers from 0 to 2147483647.
.TP
.B \-H
Precede each line with its file name, even for a single file.
When both
.B \-H
and
.B \-h
are given, the later one wins.
.TP
.B \-h
Never precede lines with file names.
.TP
.B \-n
Precede each line with its line number, starting at 1.
.TP
.B \-l
Print only the name of each file that has a selected line, and stop
reading the file at that line.
.TP
.B \-L
Print only the name of each file that has no selected line.
.TP
.B \-c
Print only the number of selected lines in each file, after the file
name and a colon when names are shown.
.TP
.B \-o
Print only the matched part of each selected line, each match on its own
line.
Empty regular expression matches are not printed.
See
.B NOTES
for how this differs from GNU
.BR grep .
.TP
.BR \-q ", " \-\-quiet
Print no lines, and exit with status 0 at the first selected line, even
if there were errors.
With
.B \-c
or
.BR \-L ,
files with no selected line are still reported until then.
.TP
.B \-s
Do not report files that cannot be opened.
The exit status still reports the error.
Directories that
.B \-r
or
.B \-R
cannot read are reported anyway, and do not affect the exit status.
.TP
.B \-r
Search the files under each directory operand, recursively.
Below the operands, symbolic links and junctions are followed if they
lead to files and skipped if they lead to directories.
File operands are searched as usual, and with no
.I file
operands
.B grep
reads standard input.
Without
.B \-r
or
.BR \-R ,
a directory operand is an error.
.TP
.B \-R
Like
.BR \-r ,
but also follow symbolic links and junctions to directories.
.TP
.B \-v
Select the lines that do not match.
.TP
.B \-i
Ignore case, for ASCII letters only.
.TP
.B \-w
Select a line only if a match is a whole word: one that is neither
preceded nor followed by an ASCII letter, digit, or underscore.
.TP
.B \-x
Select a line only if a pattern matches the whole line.
Overrides
.BR \-w .
.TP
.B \-F
Patterns are fixed strings.
Takes precedence over
.BR \-E ,
including for
.BR egrep .
.TP
.B \-E
Patterns are extended regular expressions.
.TP
.BI "\-m " num
Stop reading a file after its
.IR num th
selected line.
Trailing context is still printed after it, up to the next selected line.
With a
.I num
of 0, only the first line is read; it is never printed, but if it is
selected, it counts as such for
.BR \-c ,
.BR \-l ,
.BR \-L ,
.BR \-q ,
and the exit status.
.TP
.BI "\-A " num
Print
.I num
lines of trailing context after each selected line.
.TP
.BI "\-B " num
Print
.I num
lines of leading context before each selected line.
.TP
.BI "\-C " num
Same as
.BI "\-A " num
.BI "\-B " num
at this position, so it overrides an earlier
.B \-A
or
.BR \-B ,
and a later one overrides it.
.TP
.BI "\-e " pattern
Use
.IR pattern .
May be repeated, and combined with
.BR \-f .
.TP
.BI "\-f " patfile
Read patterns from
.IR patfile ,
one per line, ignoring a CR before each LF.
A
.I patfile
of
.B \-
is standard input.
May be repeated.
When the files hold no patterns and there is no
.BR \-e ,
no line matches.
.TP
.BR \-a ", " \-I
Accepted and ignored.
.TP
\fB\-\-color\fR[\fB=\fIwhen\fR]
Accepted and ignored.
There is no color output.
.TP
.B \-\-help
As the first argument, print a usage message on standard output and
exit with status 0.
.B egrep
and
.B fgrep
print that no help is available instead.
.SH ENVIRONMENT
.TP
.B BB_GLOBBING
Unless it is
.BR 0 ,
the C runtime replaces each argument containing wildcards with the names
of the files it matches, if any, even when the argument is quoted.
From
.BR cmd.exe ,
a pattern such as
.B \(dqa*\(dq
may become a list of files.
When it is unset, busybox sets it to
.B 0
for its child processes, so commands run from the busybox shell, which
expands wildcards itself, receive their arguments as given.
.TP
.BR LC_ALL ", " LC_CTYPE ", " LANG
If the first of these that is set has the value
.BR C ,
output to a console is not converted, as described in
.BR NOTES ,
and the console interprets the bytes in its own code page.
.TP
.B POSIXLY_CORRECT
If set, options must come before operands.
.PP
Variables that affect every busybox applet, such as
.B BB_TERMINAL_MODE
for console output, are described in
.BR busybox (1).
.SH EXIT STATUS
.TP
.B 0
A line was selected, or with
.BR \-L ,
a file name was printed.
.TP
.B 1
No line was selected.
.TP
.B 2
An error: an invalid option or argument, an invalid regular expression,
a pattern file that cannot be read, or a file that cannot be opened,
even if lines were selected in others.
With
.BR \-q ,
a selected line ends the search with status 0 regardless of earlier
errors.
.TP
.B 141
Standard output is a pipe that was closed while
.B grep
was writing to it.
Output is buffered, and a closed pipe found only by the final flush at
exit, as with short output, is ignored.
.SH EXAMPLES
Show where
.B main
is defined, with line numbers and a line of context:
.PP
.RS 4
.EX
$ grep \-n \-C1 main hello.c
2\-
3:int main(void)
4\-{
.EE
.RE
.PP
Search a source tree for the whole word
.BR count :
.PP
.RS 4
.EX
$ grep \-rnw count src
src/count.c:1:static int count;
src/count.c:5:    return count++;
.EE
.RE
.PP
Find errors and warnings in a build log, using an extended regular
expression, then the same with a basic one:
.PP
.RS 4
.EX
$ grep \-E \(aqerror|warning\(aq build.log
$ grep \(aqerror\e|warning\(aq build.log
.EE
.RE
.PP
Search for a pattern that begins with a dash:
.PP
.RS 4
.EX
$ grep \-e \-Wall Makefile
.EE
.RE
.PP
List the C files that do not include
.IR stdio.h :
.PP
.RS 4
.EX
$ grep \-L \(aq#include <stdio.h>\(aq *.c
.EE
.RE
.PP
Test for a match in a script:
.PP
.RS 4
.EX
$ grep \-q \(aq^CFLAGS\(aq Makefile && echo configured
.EE
.RE
.SH NOTES
Differences from GNU
.BR grep :
.IP \(bu 2
.B \-r
with no
.I file
operands reads standard input, not the current directory.
.IP \(bu 2
There is no binary file detection, a NUL byte ends a line, and
.B \-a
and
.B \-I
have no effect.
.IP \(bu 2
With
.BR \-F ,
.B \-o
prints the pattern as given, once per line, however often and in
whatever case it occurs.
With several patterns,
.B \-o
prints the matches of only one of them on each line; join them with
.B \e|
or
.B |
instead.
With
.BR \-w ,
.B \-o
prints every match on a selected line, whole word or not, and prints the
wrong text when the first match on the line is not a whole word.
Each later match on a line is found as if the line began where the
previous match ended, which changes the result of
.BR \eb ,
.BR \eB ,
.BR \e< ,
and
.BR \e> .
Context lines are printed whole, and with
.BR \-v ,
.B \-o
prints only the file name and line number prefixes of context lines,
without their text or a newline.
.IP \(bu 2
A regular expression is compiled when first needed, so an invalid one is
not reported if no input is read.
.IP \(bu 2
Read errors end a file silently.
.IP \(bu 2
An empty
.B \-f
file, with no other patterns, selects no line, except that with
.B \-F
and
.B \-x
it selects every line except one reading
.BR .* ,
and with
.B \-w
but neither
.B \-F
nor
.BR \-x ,
every line that begins with an ASCII letter, digit, or underscore.
.IP \(bu 2
With
.B \-w
but not
.BR \-x ,
an empty pattern, such as a blank line in a
.B \-f
file, matches every line if
.B \-F
is given, and otherwise only lines that are empty or begin with a
character other than an ASCII letter, digit, or underscore.
.IP \(bu 2
GNU options not listed above, such as
.BR \-P ,
.BR \-z ,
.BR \-\-include ,
and long forms such as
.BR \-\-ignore\-case ,
are invalid.
.PP
Output lines end with LF, not CRLF, even when the input has CRLF line
endings.
.PP
Patterns and input are compared as bytes, so their encodings must agree.
On Windows 10 version 1903 and later, busybox runs with UTF-8 as its code
page, and patterns and file names from the command line are UTF-8.
On earlier versions they are in the ANSI code page.
When standard output is a console, output is converted from that code
page to the console's for display, and escape sequences in it take
effect.
.PP
File operands may use drive letters and backslashes, and are printed as
given.
.B \-r
joins the names it finds with
.BR / ,
so
.B grep \-r x \(aqC:\esrc\(aq
prints names such as
.IR C:\esrc/main.c .
A bare drive such as
.I C:
names the current directory of that drive, and
.B \-r
prints names such as
.I C:name
for the files below it.
The device names
.IR /dev/null ,
.IR /dev/stdin ,
and
.I /dev/tty
(the console) work as on Unix.
.SH SEE ALSO
.BR busybox (1),
.BR find (1),
.BR less (1),
.BR sed (1),
.BR sh (1),
.BR xargs (1)