Path: utzoo!attcan!uunet!decwrl!elroy.jpl.nasa.gov!jpl-devvax!lwall
From: lwall@jpl-devvax.JPL.NASA.GOV (Larry Wall)
Newsgroups: comp.lang.perl
Subject: Re: What I want in regular expressions
Message-ID: <8513@jpl-devvax.JPL.NASA.GOV>
Date: 27 Jun 90 17:31:43 GMT
References: <1990Jun27.144859.15573@uvaarpa.Virginia.EDU>
Reply-To: lwall@jpl-devvax.JPL.NASA.GOV (Larry Wall)
Organization: Jet Propulsion Laboratory, Pasadena, CA
Lines: 49

In article <1990Jun27.144859.15573@uvaarpa.Virginia.EDU> worley@compass.com writes:
: There are a number of features that could be put into regexp matching
: that would make life easier:
: 
: 	a \- b
: 
: Match any string which matches a but is not matched (in its entirity)
: by b.
: 
: 	\- b
: 
: Match the null string if what follows this position is not matched by
: b.
: 
: 	a \: b \; c
: 
: Match b but only if its left context is matched by a and its right
: context is matched by c.  This would be useful for simplifying s/ / /
: constructions:
: 
: 	s/(a)b(c)/\1d\2/
: 
: becomes
: 
: 	s/a\:b\;c/d/
: 
: Probably, a should be included in $` and c in $'.
: 
: How hard would this be to do?

I don't think "hard" is the quite right question.  I have three problems with
this:

	1) It's non-standard, so people won't think to use it.  Hence
	    it would tend to be excess baggage.

	2) After scratching my head for some time, I can't think what
	    all the ramifications of the first two constructs might be.
	    And I think there are a certain number of people in the world
	    who are more confusable than me.

	3) I am opposed to backwhacking punctuation to make metacharacters,
	    for readability and quotability reasons.

But it would be nice to be able to do boolean lookahead.  Someday I might
make it possible to call subroutines within a pattern to allow this
sort of semantics.  It looks cleaner to me.

Larry