xref: /csrg-svn/usr.bin/sed/POSIX (revision 56067)

#	@(#)POSIX	5.7 (Berkeley) 08/27/92

Comments on the IEEE P1003.2 Draft 12
     Part 2: Shell and Utilities
  Section 4.55: sed - Stream editor

Diomidis Spinellis <dds@doc.ic.ac.uk>
Keith Bostic <bostic@cs.berkeley.edu>

In the following paragraphs, "wrong" usually means "inconsistent with
historic practice", as most of the following comments refer to
undocumented inconsistencies between the historical versions of sed and
the POSIX 1003.2 standard.  All the comments are notes taken while
implementing a POSIX-compatible version of sed, and should not be
interpreted as official opinions or criticism towards the POSIX committee.
All uses of "POSIX" refer to section 4.55, Draft 12 of POSIX 1003.2.

 1.	Historic implementations of sed strip the text arguments of the
	a, c and i commands of their initial blanks, i.e.

	#!/bin/sed -f
	a\
		foo\
		bar

	produces:

	foo
	bar

	POSIX does not specify this behavior.  This implementation follows
	historic practice.

 2.	Historical versions of sed required that the w flag be the last
	flag to an s command as it takes an additional argument.  This
	is obvious, but not specified in POSIX.

 3.	Historical versions of sed required that whitespace follow a w
	flag to an s command.  This is not specified in POSIX.  This
	implementation permits whitespace but does not require it.

 4.	Historical versions of sed permitted any number of whitespace
	characters to follow the w command.  This is not specified in
	POSIX.  This implementation permits whitespace but does not
	require it.

 5.	The rule for the l command differs from historic practice.  Table
	2-15 includes the various ANSI C escape sequences, including \\
	for backslash.  Some historical versions of sed displayed two
	digit octal numbers, too, not three as specified by POSIX.  POSIX
	is a cleanup, and is followed by this implementation.

 6.	The POSIX specification for ! does not specify that for a single
	command the command must not contain an address specification
	whereas the command list can contain address specifications.  The
	specification for ! implies that "3!/hello/p" works, and it never
	has, historically.  Note,

		3!{
			/hello/p
		}

	does work.

 7.	POSIX does not specify what happens with consecutive ! commands
	(e.g. /foo/!!!p).  Historic implementations allow any number of
	!'s without changing the behaviour.  (It seems logical that each
	one might reverse the behaviour.)  This implementation follows
	historic practice.

 8.	Historic versions of sed permitted commands to be separated
	by semi-colons, e.g. 'sed -ne '1p;2p;3q' printed the first
	three lines of a file.  This is not specified by POSIX.
	Note, the ; command separator is not allowed for the commands
	a, c, i, w, r, :, b, t, # and at the end of a w flag in the s
	command.  This implementation follows historic practice and
	implements the ; separator.

 9.	Historic versions of sed terminated the script if EOF was reached
	during the execution of the 'n' command, i.e.:

	sed -e '
	n
	i\
	hello
	' </dev/null

	did not produce any output.  POSIX does not specify this behavior.
	This implementation follows historic practice.

10.	POSIX does not specify that the q command causes all lines that
	have been appended to be output and that the pattern space is
	printed before exiting.  This implementation follows historic
	practice.

11.	Historical implementations do not output the change text of a c
	command in the case of an address range whose first line number
	is greater than the second (e.g. 3,1).  POSIX requires that the
	text be output.  Since the historic behavior doesn't seem to have
	any particular purpose, this implementation follows the POSIX
	behavior.

12.	POSIX does not specify whether address ranges are checked and
	reset if a command is not executed due to a jump.  The following
	program will behave in different ways depending on whether the
	'c' command is triggered at the third line, i.e. will the text
	be output even though that command is never logically encountered
	in the script by line 3.

	2,4b
	1,3c\
		text

	Historic implementations of sed, for the above example, would
	never output the test.  There was a bug, however, that if the
	"1,3" was replaced by a RE address they would output the text
	after the branch no longer applied, but would then quit without
	further processing.  For example:

	2,4b
	/one/,/three/c\
		text

	with the input:

		one
		two
		three
		four
		five
		six

	would output:

		two
		three
		four
		text

	This implementation never outputs the text, for either example.
	This is based on the belief that it would be reasonable to want
	to output some text if the pattern /one/,/three/ occurs but only
	if it occurs outside of the range of lines 2 to 4.

13.	Historical implementations allow an output suppressing #n at the
	beginning of -e arguments as well as in a script file.  POSIX
	does not specify this.  This implementation follows historical
	practice.

14.	POSIX does not explicitly specify how sed behaves if no script is
	specified.  Since the sed Synopsis permits this form of the command,
	and the language in the Description section states that the input
	is output, it seems reasonable that it behave like the cat(1)
	command.  Historic sed implementations behave differently for "ls |
	sed", where they produce no output, and "ls | sed -e#", where they
	behave like cat.  This implementation behaves like cat in both cases.

15.	The POSIX requirement to open all wfiles from the beginning makes
	sed behave nonintuitively when the w commands are preceded by
	addresses or are within conditional blocks.  This implementation
	follows historic practice and POSIX, by default, and provides the
	-a option which opens the files only when they are needed.

16.	POSIX does not specify how escape sequences other than \n and \D
	(where D is the delimiter character) are to be treated.  This is
	reasonable, however, it also doesn't state that the backslash is
	to be discarded from the output regardless.  A strict reading of
	POSIX would be that "echo xyz | sed s/./\a" would display "\ayz".
	As historic sed implementations always discarded the backslash,
	this implementation does as well.

17.	POSIX specifies that an address can be "empty".  This implies
	that constructs like ",d" or "1,d" and ",5d" are allowed.  This
	is not true for historic implementations or this implementation
	of sed.

18.	The b t and : commands are documented in POSIX to ignore leading
	white space, but no mention is made of trailing white space.
	Historic implementations of sed assigned different locations to
	the labels "x" and "x ".  This is not useful, and leads to subtle
	programming errors, but it is historic practice and changing it
	could theoretically break working scripts.  This implementation
	follows historic practice.

19.	Although POSIX specifies that reading from files that do not exist
	from within the script must not terminate the script, it does not
	specify what happens if a write command fails.  Historic practice
	is to fail immediately if the file cannot be opened or written.
	This implementation follows historic practice.

20.	Historic practice is that the \n construct can be used for either
	string1 or string2 of the y command.  This is not specified by
	POSIX.  This implementation follows historic practice.

21.	POSIX does not specify if the "Nth occurrence" of an RE in a
	substitute command is an overlapping or a non-overlapping one,
	i.e. what is the result of s/a*/A/2 on the pattern "aaaaa aaaaa".
	Historical practice is to drop core or only do non-overlapping
	RE's.  This implementation only does non-overlapping RE's.

22.	Historic implementations of sed ignore the RE delimiter characters
	within character classes.  This is not specified in POSIX.  This
	implementation follows historic practice.

23.	Historic implementations handle empty RE's in a special way: the
	empty RE is interpreted as if it were the last RE encountered,
	whether in an address or elsewhere.  POSIX does not document this
	behavior.  For example the command:

		sed -e /abc/s//XXX/

	substitutes XXX for the pattern abc.  The semantics of "the last
	RE" can be defined in two different ways:

	1. The last RE encountered when compiling (lexical/static scope).
	2. The last RE encountered while running (dynamic scope).

	While many historical implementations fail on programs depending
	on scope differences, the SunOS version exhibited dynamic scope
	behaviour.  This implementation also uses does dynamic scoping, as
	this seems the most useful and in order to remain consistent with
	historical practice.