Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!swrinde!elroy.jpl.nasa.gov!jpl-devvax!lwall
From: lwall@jpl-devvax.jpl.nasa.gov (Larry Wall)
Newsgroups: comp.lang.perl
Subject: Re: global pattern matching question
Message-ID: <1991May6.233109.12524@jpl-devvax.jpl.nasa.gov>
Date: 6 May 91 23:31:09 GMT
References: <JES.91May3121528@mbio.med.upenn.edu> <1991May4.010307.11792@jpl-devvax.jpl.nasa.gov> <JES.91May5190550@mbio.med.upenn.edu>
Reply-To: lwall@jpl-devvax.JPL.NASA.GOV (Larry Wall)
Distribution: comp
Organization: Jet Propulsion Laboratory, Pasadena, CA
Lines: 53

In article <JES.91May5190550@mbio.med.upenn.edu> jes@mbio.med.upenn.edu (Joe Smith) writes:
: The book says that $` and $' aren't generated unless they're
: mentioned.  In an expression like length $`, does perl actually
: generate a copy of this piece of the original string just to determine
: the match position, or is the length just calculated from the match
: position?

Unfortunately, "length $`" counts as a mention currently.  I've wanted
to optimize that out for some time, but haven't collected enough of
those little round tuits.

: > An option to s/// not to do replacement would be interesting, though
: 
: And to think I've always wondered how such wonderfully
: convoluted language design happens...

Yah.

: > there might be a better way--maybe an initial offset for m//, or
: > allowing patterns in index().
: 
: This seems more direct, but the ability to bind an action to the match
: is also very important.  Perhaps access to a primitive pattern
: matching function (the way splice is related to push, pop, etc.) would
: be reasonable:
: 
:    $options = "oi";
: 
:    for ($pos = 0;
: 	   $matched = regexp( $pattern, $seq, $pos, $options );
: 	   $pos += $matched ) {
: 		   push( @positions, $pos );
: 		   push( @sites, subst($seq, $pos, $matched ));
:    }
: 
: I think you could replicate the match and substitute capabilities with
: that (subst($seq, $pos, $matched ) = "your replacement here..."), with
: arbitrary actions on a match and without a lot of extra string copies.
: It'll never fit on one line, but I could live with that.

I'm thinking more along the lines of the g// operator in ed:

	while ($foo =~ g/pat/) {
	    push(@positions, length $`);
	    push(@sites, $&);
	}

g// would have to have its own built in state.  I'm not sure what it
should do if $foo changes...

Perhaps it ought to be m/pat/g instead.

Larry