in reply to Re^5: substr question
in thread substr question

I generally prefer to break things into more verbose code to make life easier 2 years down the road (or next week)

How is trying to understand 50 instructions easier than understanding 5. Your code is so complex it would take quite some time for me to understand it now, in weeks, in years.

experience said my approach should be faster since more but easier (for Perl) regexes are typically faster than one complex one.

Both of your regex are more complex than mine. (All three read linearly, but yours are longer.) Plus you have numerous additional Perl ops. It makes no sense for your code to be faster when the regex are executed. And it's not. You probably didn't take into account that your solution modifies the input.

('x' x 90).' '.('x' x 10) Rate stevenmay ikegami2 ikegami1 stevenmay 256955/s -- -35% -48% ikegami2 394568/s 54% -- -19% ikegami1 490119/s 91% 24% -- ('x' x 10).' '.('x' x 90) Rate stevenmay ikegami2 ikegami1 stevenmay 102399/s -- -13% -19% ikegami2 118154/s 15% -- -7% ikegami1 126866/s 24% 7% --

In the case where the string is shorter than 100, having the check makes it faster. You can always add that to mine. That's what ikegami2 is.

('x' x 99) Rate ikegami1 stevenmay ikegami2 ikegami1 771011/s -- -81% -83% stevenmay 4071762/s 428% -- -12% ikegami2 4620312/s 499% 13% --

Benchmark code:

use strict; use warnings; use Benchmark qw( cmpthese ); my %tests = ( stevenmay => <<'__EOI__', my $string = $input; if ( $string and length $string > 100 ){ $string = substr( $string, 0, 100); my ($tmp) = $string =~ /(.+)\s.*?$/s; # last space if possible $tmp or ($tmp) = $string =~ /(.+)\W.*?$/s; # bust on last non-word $tmp and $string = $tmp; } __EOI__ ikegami1 => <<'__EOI__', my ($string) = $input =~ /^(.{0,100})(?!\S)/s; __EOI__ ikegami2 => <<'__EOI__', my $string = $input; if (length($string) > 100) { $string =~ s/^(.{0,100})(?!\S)\K.*//s } __EOI__ ); $_ = 'use strict; use warnings; our $input; '.$_.' 1' for values(%tests); { local our $input = ('x' x 90).' '.('x' x 10); cmpthese(-1, \%tests); } { local our $input = ('x' x 10).' '.('x' x 90); cmpthese(-1, \%tests); } { local our $input = ('x' x 99); cmpthese(-1, \%tests); }

Replies are listed 'Best First'.
Re^7: substr question
by stevenmay (Initiate) on Jun 19, 2010 at 16:25 UTC

    'Your code is so complex it would take quite some time for me to understand it '

    Certainly it depends on what a person is used to. Let's agree we approach things differently at this point.

    Note that I *am* concerned about generating code that is opaque to others, and am taking your comment seriously.

    The interesting bit is your benchmark results.... I went back to look at the benchmark I ran last night.

    Oops.

    Typo.

    I stand corrected, my code is slower.
    I'd set up a test string that was less than 100 characters. So the extra ops were not even looked at and once that was corrected, my code was approximately 3 times slower. Darn.

    I do thank you for the time you've spent on me. While you might think it wasted, I certainly do not.

    \s