how to speed up a string-substitution loop?
in [Perl]
|
From: Adam Kellas on 2 Mar 2010 23:38 On Mar 2, 1:07 pm, J rgen Exner <jurge...(a)hotmail.com> wrote: > BTW: $substr is an awful name considering there is a function substr() Granted, changed. > and capitalized names ($Variables) normally indicate file handles. OT I know, but I thought ALLCAPS names indicated file handles (as in STDOUT) and CamelCase indicated global (or more properly "widely scoped") variables? AK
From: Uri Guttman on 3 Mar 2010 00:12 >>>>> "AK" == Adam Kellas <adam.kellas(a)gmail.com> writes: AK> On Mar 2, 8:04�pm, "Uri Guttman" <u...(a)StemSystems.com> wrote: >> why the strange use of % for a delimiter? it is impossible to >> read. actually one of the worst choices you can make IMO. besides you >> don't need to do that as / is fine since you have no / in your regex. AK> Style ... aesthetics ... religion ... I decided years ago that my AK> personal style guide would be to always use m%% and s%%%. It may be AK> ugly but it's consistent; I can always find patterns in my code by AK> searching for m% or s%. Try that with //! The other problem with // is AK> that half the time there are slashes in the pattern, and often the AK> pattern changes in such a way that you either can or have to change AK> the delimiter. I've never regretted adding this convention to my style AK> guide, though I'm sure a case could be made for preferring some other AK> inert character like s,,,. this isn't religion. there are plenty of religious wars on the net. this is about quality code. readability matters. it makes your code easier to understand, maintain, modify, etc. you may think this is cute and your style but most others don't. someone else posted the same thing. as for having / in the data, there is no way it is used that often. and regexes don't usually change on the fly too often. finally the best (and this comes from damian conway, among many others) alternate delimiter is {}. it allows nested pairs of {} inside without escaping, it is easy to see, it doesn't hide the regex itself. and you can even split the two parts of s{} {} with spaces and it works. >> � AK> � � for (keys %MakeVars) { >> >> use a named variable for loops. it is cleaner and easier to read. $_ can >> be easily clobbered or some action at a distance can happen. my vars are >> lexical to the loop and much safer AK> Use of $_ was deliberate here because I gather it's marginally faster AK> and I was trying to tune the loop for speed. But in general I agree AK> that named variables are preferable. not fast enough to be worth it. we are talking minute speedups and your code has other issues. readibility and safety easily trump the miniscule speedup you may get in this case. >> � AK> � � � �my $name = $MakeVars{$_}; >> � AK> � � � �$text =~ s%$_%$name%g; >> >> again with the % delimiter. i would go blind reading that. and there is >> no need for the temp $name as you can put that hash lookup directly in >> the replacement string. AK> Style ... aesthetics ... religion ... wrong, wronger and wrongest. uri -- Uri Guttman ------ uri(a)stemsystems.com -------- http://www.sysarch.com -- ----- Perl Code Review , Architecture, Development, Training, Support ------ --------- Gourmet Hot Cocoa Mix ---- http://bestfriendscocoa.com ---------
From: C.DeRykus on 3 Mar 2010 01:24 On Mar 2, 8:22 pm, Adam Kellas <adam.kel...(a)gmail.com> wrote: > On Mar 2, 8:04 pm, "Uri Guttman" <u...(a)StemSystems.com> wrote: > > > why the strange use of % for a delimiter? it is impossible to > > read. actually one of the worst choices you can make IMO. besides you > > don't need to do that as / is fine since you have no / in your regex. > > Style ... aesthetics ... religion ... I decided years ago that my > personal style guide would be to always use m%% and s%%%. It may be > ... You know of course religious threads will go on and on :) The non-denominational /x switch can help with 's' pattern readability: $text =~ s% $_ %$name%gx; And bracketing delimiters such as (),<>,{},[] can help in seeing pattern/replacement parts if you're ever tempted to stray from The True Way: $test =~ s{ $_ } {$name}gx; and can even be on multiple lines: $test =~ s[ really long pattern ] [some_replacement]gx; -- Charles DeRykus
From: Adam Kellas on 3 Mar 2010 09:29 On Mar 3, 12:12 am, "Uri Guttman" <u...(a)StemSystems.com> wrote: > this isn't religion. there are plenty of religious wars on the net. this > is about quality code. readability matters. it makes your code easier to > understand, maintain, modify, etc. you may think this is cute and your > style but most others don't. someone else posted the same thing. as for > having / in the data, there is no way it is used that often. and regexes > don't usually change on the fly too often. finally the best (and this > comes from damian conway, among many others) alternate delimiter is > {}. it allows nested pairs of {} inside without escaping, it is easy to > see, it doesn't hide the regex itself. and you can even split the two > parts of s{} {} with spaces and it works. I don't consider it cute, I consider it a convention. The value of a convention or standard is independent of its objective right-ness. Perhaps if I could go back in time I'd make some different choices, but wouldn't we all? Though how you can argue that % is ugly while / is not while explicitly rejecting the notion that it's an aesthetic disagreement - and that "quality code" belongs to your side - is beyond me. You might as well argue that yellow is uglier than blue. I don't object to this line of discussion though. Perhaps you will save some naive young person from the horrors of addiction to percent signs. For me it's 15 years and thousands of lines of maintained code too late. > AK> Use of $_ was deliberate here because I gather it's marginally faster > AK> and I was trying to tune the loop for speed. But in general I agree > AK> that named variables are preferable. > > not fast enough to be worth it. we are talking minute speedups and your > code has other issues. readibility and safety easily trump the miniscule > speedup you may get in this case. Really, we have no disagreement here. It's just that when one is trying desperately to speed something up it's reasonable to try everything you know of. What I posted was not a published, supported piece of code, it was the result of a tuning exercise. AK
From: Uri Guttman on 3 Mar 2010 10:21
>>>>> "AK" == Adam Kellas <adam.kellas(a)gmail.com> writes: AK> On Mar 3, 12:12�am, "Uri Guttman" <u...(a)StemSystems.com> wrote: >> this isn't religion. there are plenty of religious wars on the net. this >> is about quality code. readability matters. it makes your code easier to >> understand, maintain, modify, etc. you may think this is cute and your >> style but most others don't. someone else posted the same thing. as for >> having / in the data, there is no way it is used that often. and regexes >> don't usually change on the fly too often. finally the best (and this >> comes from damian conway, among many others) alternate delimiter is >> {}. it allows nested pairs of {} inside without escaping, it is easy to >> see, it doesn't hide the regex itself. and you can even split the two >> parts of s{} �{} with spaces and it works. AK> I don't consider it cute, I consider it a convention. The value of a AK> convention or standard is independent of its objective right-ness. AK> Perhaps if I could go back in time I'd make some different choices, AK> but wouldn't we all? Though how you can argue that % is ugly while / AK> is not while explicitly rejecting the notion that it's an aesthetic AK> disagreement - and that "quality code" belongs to your side - is AK> beyond me. You might as well argue that yellow is uglier than blue. but convention implies many people doing it. you are the only one i have ever seen (and i see tons of code) using % for a delimiter. so you are not conventional but instead out of the norm. it is your convention and others find it hard to read. you write code for others, not yourself. get over yourself with this. you are wrong in conventional ways. AK> I don't object to this line of discussion though. Perhaps you will AK> save some naive young person from the horrors of addiction to percent AK> signs. For me it's 15 years and thousands of lines of maintained code AK> too late. you can always change style. i have many times over my career as i learn better ways to code. >> � AK> Use of $_ was deliberate here because I gather it's marginally faster >> � AK> and I was trying to tune the loop for speed. But in general I agree >> � AK> that named variables are preferable. >> >> not fast enough to be worth it. we are talking minute speedups and your >> code has other issues. readibility and safety easily trump the miniscule >> speedup you may get in this case. AK> Really, we have no disagreement here. It's just that when one is AK> trying desperately to speed something up it's reasonable to try AK> everything you know of. What I posted was not a published, supported AK> piece of code, it was the result of a tuning exercise. desperation is not a reason to try everything under the sun for speedups. just applying the benchmark module is a saner way to find out what is actually faster and by how much. uri -- Uri Guttman ------ uri(a)stemsystems.com -------- http://www.sysarch.com -- ----- Perl Code Review , Architecture, Development, Training, Support ------ --------- Gourmet Hot Cocoa Mix ---- http://bestfriendscocoa.com --------- |