Re: [MacPerl] ?pattern? and reset

[Tim Dierks <timd@consensus.com>]

At 11:44 PM -0800 2/17/98, Stephan Somogyi wrote:
>I have to churn through around a gigabyte of text. After having sorted
>out the basic algorithmic issues, I am now looking to optimize.
>
>One of the more time critical operations is the dual pattern-matche
>that I perform on each line of text with the following:
>
>	unless (($curLine =~ /\[\d/) && ($curLine =~ /\"$\/\Z/))
>
>Since I really need to know only whether these patterns are in the
>string at all, I would prefer to use ?? rather than // patterns since,
>if I replace the // delimiters with ?? delimiters, I see a 14%
>performance improvement; unfortunately, the line of code also stops
>working thereafter. Just to make sure, I printed out $& after the line
>and it seemed to be matching the latter expression but not returning
>true.
>
>After some digging around, I discovered that ? patterns have the
>irksome need for a reset after every use. When I added the necessary
>resets, I found myself at 14% worse than when I started, ie 28% slower
>than with the ? patterns.

Put o on the second regexp so it won't have to be recompiled every time
just in case you change $/. ($curLine =~ /\"$\/\Z/o) (Unless you're
changing $/, that is: even then, it might be faster to have several
regexps, one for each possible value). This will be a big win.

Are you sure this is the only thing you're changing? I can't see any reason
?? should be faster than // (in fact, it should be slower).

If the second one matches less often, put it first to let you exit the &&
chain faster.

It might be faster to combine them into one regexp if you know the order:

	unless ($curLine =~/\[\d.*\"$\/\Z/o).

 - Tim

Tim Dierks - timd@consensus.com - www.consensus.com
  Director of Engineering - Consensus Development
  Developer of SSL Plus: SSL 3.0 Integration Suite



***** Want to unsubscribe from this list?
***** Send mail with body "unsubscribe" to mac-perl-request@iis.ee.ethz.ch