http://qs1969.pair.com?node_id=11115830


in reply to Re: Growing strings in search
in thread Growing strings in search

One precompiles an expression using qr// to avoid repeated compilations later. 
my $RE = qr/foo.+?bar/;
foreach (@str) { warn "Huzzah! $str" if m/$RE/; }

[download]

--
In Bob We Trust, All Others Bring Data.

Replies are listed 'Best First'.
Re^3: Growing strings in search
by AnomalousMonk (Archbishop) on Apr 21, 2020 at 00:35 UTC

    What I had in mind with my comment about the  m//o modifier is that b4swine might be receiving a complex regex and then interpolating it a la  m/...$re.../ during actual matching (as, indeed, you show in your example). This might be done if some additional match condition(s) were to be imposed on every received regex, e.g., adding a  \z anchor at the end to force end-of-string matching. Of course, I don't know that anything like this is actually happening; it was just a thought.

    For regexes interpolated into  m// matches, use of the  /o modifier seems beneficial in some cases, apparently depending on Perl version. (I know that work is being done continuously to refine and optimize regex matching.) For instance, under two older Perl versions that I have access to ATM (ActiveState 5.8.9 and Strawberry 5.14.4.1), benchmarking shows a significant benefit in the newer Perl for using  /o when interpolating into an  m// match (assuming this is a valid benchmark, and benchmarks can be tricky :). 

    use strict;
use warnings;

use Benchmark qw(cmpthese);

print "perl version $] \n";

my $rx = qr{X}xms;

cmpthese(-1, {
    'qr_bound'    => sub { 'Y' =~   $rx   ; },
    'qr_interp'   => sub { 'Y' =~ / $rx / ; },
    'qr_interp_o' => sub { 'Y' =~ / $rx /o; },
    'm_empty'     => sub { 'Y' =~ //;       },
    'm_empty_o'   => sub { 'Y' =~ //o;      },
    });

    [download]
    Output: 
    c:\@Work\Perl\monks\belg4mit>perl cmp_qr_compilation_1.pl
perl version 5.008009
                 Rate   qr_interp qr_interp_o    qr_bound     m_empty 
+ m_empty_o
qr_interp   2880477/s          --         -9%        -43%        -55% 
+      -67%
qr_interp_o 3181791/s         10%          --        -37%        -51% 
+      -64%
qr_bound    5078627/s         76%         60%          --        -21% 
+      -42%
m_empty     6452035/s        124%        103%         27%          -- 
+      -26%
m_empty_o   8775008/s        205%        176%         73%         36% 
+        --

c:\@Work\Perl\monks\belg4mit>perl cmp_qr_compilation_1.pl
perl version 5.014004
                  Rate    qr_bound  qr_interp     m_empty  m_empty_o q
+r_interp_o
qr_bound     1004609/s          --       -70%        -76%       -80%  
+      -92%
qr_interp    3378700/s        236%         --        -20%       -33%  
+      -74%
m_empty      4237856/s        322%        25%          --       -16%  
+      -68%
m_empty_o    5068698/s        405%        50%         20%         --  
+      -61%
qr_interp_o 13135439/s       1208%       289%        210%       159%  
+        --

    [download]
    The  qr// comparisons are between failing matches because that's what I imagine would happen most often in the actual application. I just threw in some successful  // null matches out of curiosity.

    Give a man a fish:  <%-{-{-{-<

[reply]
[d/l]
[select]

In Section Seekers of Perl Wisdom