search and replace regex return value

milma has asked for the wisdom of the Perl Monks concerning the following question:

Could somebody please explain why the expected message is not printed from the code below. I assumed that the regex will return true when a replacement is done, but it makes the replacement and returns false

use strict;
use warnings;
use Tk;
use Tk::BrowseEntry;

my %CFG = ( 'target' => 'foobar' );

my $main = MainWindow->new();

my $entry = $main->BrowseEntry( -variable=>\$CFG{target}, );
$entry->grid( -row=>1, -column=>1, );
my $continue = $main->Button( -command=>\&continue, -text=>'run', );
$continue->grid( -row=>2, -column=>1, );

sub continue {
    warn "$CFG{target}\n";
    if ( $CFG{target} =~ s/bar//xms ) {
        warn "expecting this message\n";
    }
    warn "$CFG{target}\n";
}
MainLoop();
[download]

Comment on search and replace regex return value Download Code

Replies are listed 'Best First'.
Re: search and replace regex return value by shmem (Chancellor) on May 07, 2017 at 10:05 UTC
Running your code with perl 5.20.2 Tk 804.032 on linux (debian 8) produces: `foobar expecting this message foo` [download] perl -le'print map{pack c,($-++?1:13)+ord}split//,ESEL'	[reply] [d/l]
Re^2: search and replace regex return value by haukex (Archbishop) on May 07, 2017 at 15:53 UTC
I can reproduce with Perl 5.24.1 and Tk 804.033 on Ubuntu. The output of the program after two button clicks is: `foobar foo foo foo` [download] Which is strange, since `... =~ s///` is supposed to return the number of substitutions made, and clearly there's one being made here. But as I said here, inspecting things with Devel::Peek shows that `$CFG{target}` appears to have `PERL_MAGIC_uvar` magic attached to it, so I guess that's causing this strange behavior (haven't yet investigated further).	[reply] [d/l] [select]
Re: Tk textvariable search and replace regex return value by Anonymous Monk on May 07, 2017 at 10:15 UTC
Could somebody please explain why the expected message is not printed from the code below. I assumed that the regex will return true when a replacement is done, but it makes the replacement and returns false I assume it has to do with the value being a textvariable Workaround is simple `my $target = $CFG{target} ; if ( $target =~ s/bar//xms ) { warn "expecting this message\n"; $CFG{target} = $target; } else { warn "not expecting this message\n"; }` [download]	[reply] [d/l]
Re: search and replace regex return value by milma (Initiate) on May 07, 2017 at 14:02 UTC
The issue is visible on perl 5.22.0, Tk 804.033 running in CentOS 6.8. Based on the fact that older version gives "correct" result I can only assume that I have a faulty version and I'll need to update to latest available and see if it fixes the issue.	[reply]
Re^2: search and replace regex return value by LanX (Saint) on May 07, 2017 at 14:15 UTC
Could it be a blessed object or something else? Probably you should check the value with Data::Dump and Devel::Peek Cheers Rolf _{(addicted to the Perl Programming Language and ☆☆☆☆ :) Je suis Charlie!}	[reply]
Re^3: search and replace regex return value by haukex (Archbishop) on May 07, 2017 at 14:47 UTC
Could it be a blessed object or something else? I think the AM post is onto something, Devel::Peek shows that `$CFG{target}` has some magic of type `PERL_MAGIC_uvar` attached to it. Not being a Tk expert I can only guess that the magic is probably necessary to bind the value to the widget, but the fact that this causes `$CFG{target} =~ ...` to behave differently than usual smells like a bug to me. But if the app needs to work on a range of Tk versions, it looks like the workaround might be necessary.	[reply] [d/l] [select]
Re^4: search and replace regex return value by LanX (Saint) on May 07, 2017 at 16:23 UTC
Re^5: search and replace regex return value by haukex (Archbishop) on May 07, 2017 at 17:00 UTC
Re^2: search and replace regex return value by milma (Initiate) on May 07, 2017 at 20:36 UTC
The problem does not exist with perl version 5.24.1 and also the workaround `my $target = $CFG{target} ;` resolves the issue in version 5.22, so I have alternative ways to resolve my issue. Thanks for your help.	[reply] [d/l]