cr> Where was I supposed to put my comments? There's not TOFU because I
cr> wasn't replying to anything in his post. Granted, I could have deleted
cr> the cc of the previous posts, but that's not something to get
cr> up-in-arms about. There was no non-linear scanning required.
it doesn't matter. your choice was wrong for this group. you defend that
with the same weak logic that you use for the regex stuff. keep trying.
cr> My last paragraph:
cr> I would wonder about memory management because the entire script is a
cr> few thousand lines long, but there aren't many global variables. And
cr> it does not slow down elsewhere. The code is almost all subroutines
cr> with 'my' variables. I assume these variables are deallocated when the
cr> subroutine is done, please correct me if I'm wrong.
cr> What else could it be?
a faulty regex? a bug in line 42? shall i whip out the venerable
PSI::ESP module and divine the answer?
cr> What of that makes you think I'm worried about the syntax of my regex?
the syntax may be fine as it compiles (or so you claim). but if you
don't know how to spot the possible issues with a regex you won't
realize that your test data isn't comprehensive enough.
cr> Okay.
cr> Though I was alarmed by John's mention of backtracking- I was afraid he
cr> meant backtracking across previous expression matches or some
cr> memory-intensive attempt at interpreter optimization by the perl folks.
cr> After reading his suggested information, I found this is not the case.cr> The 'backtracking' section of the perldoc on regex.
but without showing the regex and the slow data, you haven't proven
anything yet. you seem to keep missing this point. show the code and
data. otherwise you are really wasting all of our time. but of course
you don't mind wasting your own.
cr> Oh lets see, if my process's quantum is small and priority is low then
cr> it won't get completed very fast, now will it? If you don't understand
cr> the relationship between a process and the os running it you probably
cr> shouldn't be posting on this thread.
wow. oh well, i will have to leave you now. this was the stupidest thing
you have said so far. regexes don't massively slow down just because of
some scheduling issues. they are compute bound in almost all cases
unless they have heat death regexes where they can suck the universe
dry. but as usual without any code we will just have to take your word
that all is fine.
cr> I DON'T NEED HELP WITH THE SYNTAX OF MY REGEX. I assume shouting is
cr> warranted when you've repeated yourself so many times. I appreciate
cr> John mentioning a possible pitfall based on the particulars of Perl's
cr> matching implementation. But everything I've directly asked has dealt
cr> with the interpreter and/or how it interfaces with UNIX.
well, you know best. track all that down. see if you can figure out why
a data dependent behavior would cause major operating system changes but
not be a case of just a poorly written regex.
and you keep confusing syntax and semantics. your regex syntax is
probably correct as you claim it compiles. (hey i said that
already. maybe you will learn it this time?) on the other hand, the
semantics are surely wrong since they will slow down based on the
data.
cr> I'm glad you can be so sage and philosophical but you haven't given me
cr> anything to learn. You've just complained that I haven't supplied
cr> information that you don't need. I appreciate the fact that you are
cr> willing to run my code and input rather than give nebulous answers
cr> about the way Perl works. But it's the nebulous answers that I need.
info i don't need? the code and the data? what are you babbling about??
nebulous answers are for nebulous questions and neither have a place in
this group. if you think the problem is with unix, then go to a unix
group. admit it, you know more perl than anyone here.
unlikely but possible. and of course without seeing the reges nor the
inputs, no one but you can tell why it slows down. i suspect it is on
line 42 of the regex. remove the + (or is it a []?)
cr> Yes you can. Runtime isn't just a function of the expression. If I
cr> had a mile-long stack frame I would see a performance hit. If Perl had
cr> an ugly garbage collector and all of my previous subroutine variables
cr> hadn't been deallocated I would see a performance hit. If Perl
cr> parallelized regexes and forked many, many times I would see a
cr> performance hit. If any generally intensive script required that the
cr> coder set a lower 'nice' value I would see a performance hit.
wow x 10.
cr> This is what I asked for in the beginning, specified in all of my
cr> subsequent posts, and keeps being ignored.
cr> Oh! now I understand.
no you don't. you won't. try banging your head against your
keyboard. you will be nebulous soon.
too bad i won't be consulting for you in the future. but you can't
handle the truth.
bye bye.
uri