in reply to Re: eliminate quotes
in thread eliminate quotes

$line =~ s/"//;

Or s/"//g, if there's more than one quote per line.

tr/"//d  (note the d) would work just as well, though.

Replies are listed 'Best First'.
Re^3: eliminate quotes
by JavaFan (Canon) on Jul 29, 2008 at 12:14 UTC
    If you're using s///, and there's more than one quote per line, I'd use s/"+//g;. That ought to be somewhat faster if there are many quotes in a row. Of course, typical data doesn't have many quotes in a row, but the general principle hold, if you are deleting sub patterns, s/(?:PAT)+//g is usually faster than s/PAT//g.

      Yes, you're right, with many consecutive quotes, s/"+//g would be considerably faster than s/"//g.

      Here are a few benchmarks for those of you who like performance tuning for the sake of performance tuning... :)

      As can be seen easily, in the extreme (unrealistic) case with 999 consecutive quotes followed by a single other character, s/"+//g clearly wins, while with more realistic cases, the difference between the s/// variants becomes less pronounced, and tr/"//d is outperforming both of them.

      #!/usr/bin/perl use strict; use warnings; no warnings "numeric" ; use Benchmark "cmpthese"; my %strings = ( '999:1' => scalar ('"' x 999 . '-') x 1000, '9:1' => '"""""""""-' x 100000, '1:1' => '"-' x 500000, ); for my $test (sort {$b<=>$a} keys %strings) { my $s = $strings{$test}; printf "\n%s... ($test):\n\n", substr($s,0,30); cmpthese (1000, { 's/"//g' => sub { my $t = $s; $t =~ s/"//g; }, 's/"+//g' => sub { my $t = $s; $t =~ s/"+//g; }, 'tr/"//d' => sub { my $t = $s; $t =~ tr/"//d; }, } ); }

      ___

      $ ./700801.pl """"""""""""""""""""""""""""""... (999:1): Rate s/"//g tr/"//d s/"+//g s/"//g 4.09/s -- -98% -99% tr/"//d 212/s 5079% -- -31% s/"+//g 309/s 7445% 46% -- """""""""-"""""""""-"""""""""-... (9:1): Rate s/"//g s/"+//g tr/"//d s/"//g 4.49/s -- -85% -98% s/"+//g 30.7/s 584% -- -83% tr/"//d 181/s 3931% 490% -- "-"-"-"-"-"-"-"-"-"-"-"-"-"-"-... (1:1): Rate s/"+//g s/"//g tr/"//d s/"+//g 6.78/s -- -16% -97% s/"//g 8.05/s 19% -- -96% tr/"//d 211/s 3005% 2515% --

      (results shown for "perl, v5.8.8 built for x86_64-linux-thread-multi", but v5.10.0 behaves similarly)