NAME

comm — analyse sorted line sets

SYNOPSIS

DESCRIPTION

Analyses lines from sorted files lhs (standard input stream if "-") and rhs (standard input stream if "-"), writing them to the standard output stream, tabulated, with:

By default, therefore, the output contains lhs ∪ rhs, but columns may be suppressed, removing them entirely: if -123, only the total is written, if enabled.

Input lines in each file must be sorted according to the current locale's collation sequence (with a fall-back to byte-wise comparison on locales with an @; cf. HISTORY, Standards), and the output is ordered with respect to the same ordering.

OPTIONS

-1

Remove the first column (lines only in lhs).

-2

Remove the second column (lines only in rhs).

-3

Remove the third column (common lines).

-t, --total

Append a line listing the count of each line classification (unaffectd by -123), and "total" as the fourth column.

-z, --zero-terminated

Line separator is NUL, not newline.

-O, --output-delimiter=delim

Separate output columns with delim instead of a tab. If delim is the empty string, use a single NUL instead.

--check-order

Exit 1 as soon as an out-of-order line in either input file is detected, The default is to warn on the standard error stream and continue.

--nocheck-order

Ignore out-of-order input lines entirely.

EXIT STATUS

1 if either input file is unsorted (unless --nocheck-order) or couldn't be opened.

EXAMPLES

Separate out multiples of 2 and 3 from multiples of 6:

$ comm <(seq 0 2 12 | sort) <(seq 0 3 12 | sort)
                0
10
                12
2
        3
4
                6
8
        9

Get new entries in IEEE Std 1003.1-2008 (“POSIX.1”) XCU since Version 3 of the Single UNIX Specification (“SUSv3”), tally entries:

$ comm -13tO'  ' <(ls susv3/utilities) <(ls 9699919799/utilities)
V3_chap01.html
V3_chap02.html
V3_chap03.html
V3_chap04.html
V3_title.html
19  5  163  total

SEE ALSO

join(1), paste(1), sort(1), uniq(1), strcoll(3)

STANDARDS

Conforms to IEEE Std 1003.1-202x (“POSIX.1”), Draft 2.1 — only -123 are standard: -z, --total, --output-delimiter, --[no]check-order are extensions, originating from the GNU system; -tO are extensions.

The GNU system treats empty -O as a NUL for line output, but as the empty string for the -t line — this implementation handles it consistently as NUL.

IEEE Std 1003.1-202x (“POSIX.1”), Draft 2.1 only requires that [lr]hs be text files — do not contain NULs, and lines do not exceed LINE_MAX — common implementations in the wild conform to this limit exclusively.

HISTORY

Appeared in Version 4 AT&T UNIX as comm(I):

Version 5 AT&T UNIX removes file3 and accepts "-" to mean the standard input stream (undocumented).

Version 6 AT&T UNIX documents the "-"-as-[lr]hs behaviour.

Version 7 AT&T UNIX enforces the LINE_BUF line length maximum (now 256) instead of silently overflowing input buffers, writes errors to the standard error stream, and exits 1 for them; the presortedness requirement is now "in ASCII collating sequence", which is equivalent.

4.3BSD-Reno sees a rewrite, bumping the line limit to _BSD_LINE_MAX (2048 bytes). 4.4BSD renames that to the LINE_MAX of today.

AT&T System III UNIX exits 2 for all errors.

X/Open Portability Guide Issue 2 (“XPG2”) includes Version 7 AT&T UNIX comm verbatim, marked with an OF ("Output format incompletely specified") warning (as expected, since the manual is not a standards document and doesn't expound the output in excruciating detail).

X/Open Portability Guide Issue 3 (“XPG3”) defines locale interaction, as present-day (without fallback for locales whose collation sequence doesn't define a total ordering for all characters).

IEEE Std 1003.2-1992 (“POSIX.2”) defines the output format precisely and [lr]hs to be text files, as present-day.

IEEE Std 1003.1-202x (“POSIX.1”), Draft 2.1 requires that locales without an @ contain collation sequence with a total ordering for all characters, and defines the aforementioned fallback to be a further byte-wise comparison.