w64devkit/man/man1/sort.1

.TH SORT 1 2026-10-19 "w64devkit" "w64devkit Manual"
.SH NAME
sort \- sort lines of text
.SH SYNOPSIS
.B sort
.RI [ options ]
.RI [ file ...]
.SH DESCRIPTION
.B sort
reads the lines of every
.IR file ,
sorts them together, and writes them on standard output.
With
.BR \-c ,
it writes nothing, and only checks whether the input is sorted.
.PP
With no
.IR file ,
or for a
.I file
named
.BR \- ,
.B sort
reads standard input.
It reads all of its input into memory before it writes anything.
If a
.I file
cannot be opened,
.B sort
stops at that file, and writes nothing.
.PP
A line ends at an LF.
A CR directly before the LF is removed, so CRLF input gives LF output.
A NUL byte also ends a line, and is not written.
The last line of a file need not end in an LF.
Every output line ends in an LF.
With
.BR \-z ,
lines end at NUL bytes instead, and LF and CR are ordinary characters.
.SS Comparison
Unless a sort type option says otherwise, two lines are compared byte by
byte, as unsigned values.
The locale does not matter.
Uppercase letters sort before lowercase letters, and UTF-8 text sorts in
code point order.
.PP
At most one of
.BR \-n ,
.BR \-g ,
.BR \-h ,
.BR \-M ,
and
.B \-V
may be given, to compare numbers, month names, or versions instead.
Two of them make the first comparison fail with the message
.BR "unknown sort type" .
The options
.BR \-b ,
.BR \-d ,
.BR \-f ,
.BR \-i ,
and
.B \-r
change what is compared, or the direction.
.SS Keys
By default the whole line is compared.
Each
.B \-k
option instead compares part of the line, called a key.
The keys are compared in the order given, and the first difference
decides.
.PP
Without
.BR \-t ,
runs of white space separate fields, and the white space before a field
belongs to it.
A key that starts at field 2 therefore starts with the blanks before
it, unless
.B b
skips them.
White space is a space, TAB, LF, VT, FF, or CR.
With
.BR \-t ,
each occurrence of the separator ends a field, and belongs to neither
neighbor.
Fields can then be empty.
.PP
The argument of
.B \-k
is
.PP
.RS 4
.EX
\fIstart\fR[\fB.\fIoffset\fR][\fIopts\fR][\fB,\fIend\fR[\fB.\fIoffset\fR][\fIopts\fR]]
.EE
.RE
.PP
where each number is a decimal integer greater than 0, and offsets
count bytes.
.TP
\fIstart\fR[\fB.\fIoffset\fR]
The field where the key starts, numbered from 1.
With an
.IR offset ,
the key starts at that character of the field, counted from 1 after any
blanks skipped by
.BR b .
A start past the end of the line gives an empty key.
.TP
\fIend\fR[\fB.\fIoffset\fR]
The last field of the key, which is included whole.
Without a comma, the key runs to the end of the line.
Unlike in POSIX and GNU
.BR sort ,
an
.I offset
after
.I end
is a character position counted from the start of the line, not of the
field.
The key ends after that character, even if it lies in another field,
and keeps any white space at its end, even with
.B b
or
.BR \-b .
A key that would end before it starts is empty.
.TP
.I opts
Any of
.BR n ,
.BR g ,
.BR h ,
.BR M ,
.BR b ,
.BR r ,
.BR d ,
.BR f ,
and
.BR i ,
which act as the options of the same names, on this key only.
They apply to the whole key, whichever number they follow, except that
.B b
after
.I start
ignores white space at the start of the key, and
.B b
after
.I end
ignores it at the end.
.B V
is not allowed.
.PP
A key with
.I opts
uses only those, and none of the options that apply to the whole line.
A key without
.I opts
uses all of them.
.SS Equal lines
When all the keys of two lines compare equal, the whole lines are
compared as bytes, and
.B \-r
reverses that comparison too.
The output is then the same whatever the order of the input.
.B \-s
turns this off: equal lines keep their input order, even with
.BR \-r .
.SH OPTIONS
Options may be grouped, as in
.BR \-nr ,
and may follow the
.I file
operands.
An option's argument may be attached, as in
.BR \-k2,2n ,
or be the next argument.
.B \-\-
ends the options.
There are no long options.
.B \-o
and
.B \-t
may each be given once, and giving one again is a usage error.
.TP
.BI "\-o " file
Write the output to
.IR file ,
which is created or truncated after all the input has been read, so
.I file
can also be an input.
A
.I file
named
.B \-
is a file of that name, while
.I /dev/stdout
is standard output.
.B \-c
does not create it.
.TP
.B \-c
Check the input and write nothing.
If it is sorted, exit with status 0.
Otherwise print
.BI "Check line " N
on standard error and exit with status 1, where
.I N
is the number of the line that is followed by a line that should come
before it.
Only the first such line is reported.
Lines are numbered from 1 across all the files.
Lines with equal keys must be in the order described under
.BR "Equal lines" ,
even with
.BR \-u .
With
.BR \-u ,
identical adjacent lines are also out of order.
.TP
.B \-b
Ignore white space at the start and at the end of each key.
.TP
.B \-f
Fold lowercase ASCII letters to uppercase before comparing, so that, for
example, an underscore sorts after the letters.
.TP
.B \-i
Compare only the printable ASCII characters, 0x20 to 0x7e.
All others, such as TABs and non-ASCII bytes, are ignored.
.TP
.B \-d
Dictionary order: compare only ASCII letters, digits, and white space.
Every other character, including each byte of non-ASCII text, is ignored.
.TP
.B \-n
Compare by numeric value.
The number is read from the start of the key as C
.B atof
reads it: optional white space, a sign, decimal digits with an optional
point, and an optional exponent such as
.BR e3 .
A key with no number is 0.
Hexadecimal numbers,
.BR inf ,
and
.B nan
are not recognized, and count as 0.
.TP
.B \-g
General numeric: like
.BR \-n ,
but the number is read as C
.B strtod
reads it, so hexadecimal numbers such as
.BR 0x1f ,
.BR inf ,
and
.B nan
are recognized.
Keys with no number are equal and sort first, then NaNs, then
.BR \-inf ,
then numbers, then
.BR inf .
.TP
.B \-h
Human-readable numbers: like
.BR \-g ,
but a suffix directly after the number is compared first.
The suffixes in increasing order are none,
.B k
or
.BR K ,
then
.BR M ,
.BR G ,
.BR T ,
.BR P ,
.BR E ,
.BR Z ,
and
.BR Y .
Only
.B k
may be lowercase, and any other text after the number counts as no
suffix.
Keys with different suffixes are ordered by suffix alone, whatever the
numbers, so
.B 1500
sorts before
.BR 1K ,
and
.B 2000K
before
.BR 1M .
.TP
.B \-M
Compare by month.
The key must start with an English month name, abbreviated or in full, in
any case, after any white space.
Only that start is read, so
.B Mar
and
.B March 5
are both March.
Keys that do not start with a month name are equal, and sort before
January.
.TP
.B \-V
Compare as version strings: runs of digits are compared by numeric value,
so
.B a9
sorts before
.BR a10 .
The comparison is musl's
.BR strverscmp .
It is not allowed in a
.B \-k
option, but applies to keys without
.IR opts .
.TP
.BI "\-t " char
Use
.I char
as the field separator for
.B \-k
keys.
It must be exactly one byte, or
.B sort
prints
.B bad \-t parameter
and exits.
It has no effect without
.BR \-k .
.TP
.BI "\-k " key
Add a sort key, described under
.BR Keys .
May be repeated.
A malformed key is an error.
A number that is missing or 0 gives
.BR "bad field specification" ,
a character that is not an option letter gives
.BR "unknown key option" ,
and an option letter that keys do not allow, such as
.B V
or
.BR u ,
gives
.BR "unknown sort type" .
.TP
.B \-r
Reverse the sort order.
This applies to keys without
.IR opts ,
and to the comparison of equal lines.
A key with
.I opts
is reversed only by an
.B r
among them.
.TP
.B \-s
Stable: do not compare equal lines as bytes, so they keep their input
order, with or without
.BR \-r .
.TP
.B \-u
Write only the first line of each run of lines with equal keys.
The comparison of equal lines is not used to decide equality, so with
options such as
.BR \-f ,
.BR \-n ,
or
.BR \-k ,
lines that differ can be dropped.
The line kept is the first in sorted order: with
.BR \-s ,
the first in the input, and otherwise the smallest as bytes, or the
largest with
.BR \-r .
.TP
.B \-z
Lines end at a NUL byte in the input and the output, not at an LF.
An LF or CR inside a line is data, and no CR is removed.
.TP
\fB\-m\fR, \fB\-S\fR \fIsize\fR, \fB\-T\fR \fIdir\fR
Accepted and ignored.
There is no merging, and
.B sort
never uses temporary files.
.TP
.B \-\-help
Print a usage message on standard output and exit with status 0.
It is recognized only as the first argument.
.SH ENVIRONMENT
.TP
.B BB_GLOBBING
Unless it is
.BR 0 ,
wildcards in the arguments are expanded, as needed when run from
.BR cmd.exe .
A busybox process started with it unset sets it to
.B 0
for its children, so commands run from the kit's shell receive their
arguments as the shell expanded them.
.TP
.B POSIXLY_CORRECT
If set, the first
.I file
operand ends the options, so that later arguments such as
.B \-r
are file names.
.TP
.BR LC_ALL ", " LC_CTYPE ", " LANG
The order of the output does not depend on them.
If the first of these that is set has the value
.BR C ,
output to a console is not converted, and the console interprets the
bytes in its own code page.
.PP
Variables that affect every busybox applet are described in
.BR busybox (1).
.SH EXIT STATUS
.TP
.B 0
Success, or with
.BR \-c ,
the input is sorted.
.TP
.B 1
With
.BR \-c ,
the input is not sorted.
.TP
.B 2
An error: a usage error, such as an unknown option, a missing argument,
or a repeated
.B \-o
or
.BR \-t ;
a bad
.B \-t
or
.BR \-k ;
more than one sort type; an input file that could not be opened or is a
directory; an output file that could not be opened; a failure writing
standard output; or memory that ran out.
.TP
.B 141
Standard output is a pipe whose reader has exited.
.PP
Errors are reported on standard error, beginning with
.BR sort: .
A usage error also prints the usage message.
.SH EXAMPLES
Sort a file:
.PP
.RS 4
.EX
$ sort names.txt
.EE
.RE
.PP
List the largest numbers first:
.PP
.RS 4
.EX
$ sort \-rn sizes.txt
.EE
.RE
.PP
Sort a comma-separated file by its third column as a number, and then by
its first column as text:
.PP
.RS 4
.EX
$ sort \-t, \-k3,3n \-k1,1 data.csv
.EE
.RE
.PP
Sort a file in place:
.PP
.RS 4
.EX
$ sort \-o list.txt list.txt
.EE
.RE
.PP
Rank the words in a file by how often they occur:
.PP
.RS 4
.EX
$ sort words.txt | uniq \-c | sort \-rn
.EE
.RE
.PP
List the sizes of files and directories, smallest first:
.PP
.RS 4
.EX
$ du \-sh * | sort \-h
.EE
.RE
.PP
Sort file names that may contain any character, as NUL-terminated
strings:
.PP
.RS 4
.EX
$ find . \-name \(aq*.c\(aq \-print0 | sort \-z | xargs \-0 wc \-l
.EE
.RE
.PP
Check that a file is sorted, as
.BR comm (1)
and
.BR join (1)
need:
.PP
.RS 4
.EX
$ sort \-c list.txt && echo sorted
.EE
.RE
.SH NOTES
Differences from GNU
.BR sort :
.IP \(bu 2
Lines are compared as bytes in every locale.
.BR \-f ,
.BR \-d ,
.BR \-i ,
and
.B \-M
know only ASCII and English.
.IP \(bu 2
There are no long options.
Options not listed above, such as
.B \-C
and
.BR \-R ,
are invalid.
.IP \(bu 2
The character offset in the end of a key counts from the start of the
line, not of the field, and may not be 0.
.IP \(bu 2
.B \-n
reads exponents, which GNU
.B sort
leaves to
.BR \-g .
.IP \(bu 2
.B \-c
prints only a line number, which is that of the first line of the pair
that is out of order.
.IP \(bu 2
The sort is done in memory with no temporary files, so the input must
fit in memory.
.PP
Do not combine
.B \-c
with
.BR \-s .
When two lines have equal keys,
.B \-c
then reads position data that is stored only when sorting, and can
report sorted input as unsorted.
.PP
Output lines end with LF, not CRLF.
A UTF-8 byte order mark is not removed, and is part of the first line of
its file.
.PP
Each
.I file
may be a Windows path, with a drive letter and either kind of slash, or a
busybox-w32 device name such as
.IR /dev/null .
On Windows 10 version 1903 or later, file names may contain any Unicode
characters, and on older systems only those of the system code page.
.SH SEE ALSO
.BR busybox (1),
.BR comm (1),
.BR cut (1),
.BR join (1),
.BR uniq (1)