Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!agate!ucbvax!pasteur!dog.ee.lbl.gov!elf.ee.lbl.gov!torek
From: torek@elf.ee.lbl.gov (Chris Torek)
Newsgroups: comp.editors
Subject: Re: vi and emacs
Message-ID: <11036@dog.ee.lbl.gov>
Date: 17 Mar 91 21:32:59 GMT
References: <1991Mar06.150359.13516@chinet.chi.il.us> <1991Mar7.214419.17515@m.cs.uiuc.edu> <1991Mar07.232206.8438@convex.com> <1991Mar8.164415.14087@alchemy.chem.utoronto.ca>
Reply-To: torek@elf.ee.lbl.gov (Chris Torek)
Organization: Lawrence Berkeley Laboratory, Berkeley
Lines: 53
X-Local-Date: Sun, 17 Mar 91 13:32:59 PST

In article <1991Mar8.164415.14087@alchemy.chem.utoronto.ca>
mroussel@alchemy.chem.utoronto.ca (Marc Roussel) writes:
>Anil's point was that many Unix utilities use different flavours of
>regular expressions.  I for one find that to be a nuisance too.  Why
>should grep use different syntax for its regular expressions than ex?

This is fixed in 10th Edition Unix.  It is just an accident of history.
There *is* a reason for some difference, actually:  Editors must have
a way of marking matches for substitutions, while pure matchers (grep)
do not need this.

Things might be in much better shape if Ken Thompson had originally
designed `ed' with a single special character that introduced all
regular expression metasequences; then existing programs (and people!)
that use the various RE matchers would already quote that character and
it would not be such a problem to change them all to have some new
feature.  For instance, if `_' were the magic character, then:

	grep stdio.h *.c

would do what you meant; you would have to ask for

	grep '#_ _*include_ _*<_._*>'

to match `#, followed by any amount of space, followed by include,
followed by any amount of space, followed by <, followed by any amount
of anything except newline, followed by >'.  Here `_ ' means `space or
tab' and is just shorthand for _[   _].  `Remembering' for replacements
could be done with _(foo_) and back-references with _1 through _9.  You
could also get rid of the positional requirements for - and ^ within
character classes (I got rid of the one for ] already, above, by using
_]) by using _- and _^:

	_[^a_-zA_-Z-_]

would mean `class: caret, or a through z, or A through Z, or hyphen',
which matches things like `word' or `hyphenated-word' and includes words
spelled `r^ole'.  To match everything but that, throw _^ in somewhere.

(Now someone will argue for something other than `_'... :-)  Backslash
would be good, but the shell uses it; that is why I picked _ here.)

(Actually, if I were designing an RE syntax, I think I would make Kleene
`*' matching use a prefix, not a postfix; it means something like

	foo\* bar

would match `foobar' and `foo   bar'.  I think of * as `zero or more',
so this means "foo, zero or more o's, bar".  \+X for 1 or more X, and
\<m,n>X for `between m and n inclusive' X's, would also be useful.)
-- 
In-Real-Life: Chris Torek, Lawrence Berkeley Lab CSE/EE (+1 415 486 5427)
Berkeley, CA		Domain:	torek@ee.lbl.gov