Tracing warnings: how to show the regexp values?

[nikolay]

In the script

#!/usr/bin/perl

use warnings;
use strict;


my $sod='ip6 is used for...';
my $i=0;
my @baza_tek=qw( (?^u:ip(\d)) "internet_protocol-$1$2" );

$sod=~s#$baza_tek[$i]#$baza_tek[$i+1]#gee;

i have such an error on execution:

Use of uninitialized value $2 in concatenation (.) or string at...

My question is,

How to tell PERL to show the values of current baza_tek[$i] or $baza_tek[$i+1] at the time, when the error occured?

[LanX]

From perlre#Extended-Patterns

> "(?^aluimsx:pattern)"

This is for clustering, not capturing ; it groups subexpressions like "()", but doesn't make backreferences as "()" does.

[1nickt]

You should always check to see if your capture matched anything before using it.

if ( $2 ) {
    ...
} else {
    ...
}

Edit: ... By which I do not mean to suggest that you should always use an explicit if conditional: as my learned colleague explains below, there are any number of reasons you might get an undefined value. The point is your code should be ready for it, whether with an explicit test, no warnings, or howsoever you deem best.

[LanX]

I dare to disagree for global substitutions s///g

Either it doesn't match, so the error can't occur with a tested regex!

Or one used or clauses or conditional matching like in /(a)|(b)(c)/ then having an undefined $2 is intended behaviour.

in the latter case silencing the warning of an undefined value becoming an empty string is the normal approach. ( no warnings qw/uninitialized/ )

[nikolay]

Thank you. But i do not know exactly the number of the variables that will be used. I have brought here just a simple example.

[LanX]

I order to capture a warning for debugging, you could (locally with local ) change sigwarn

  DB<120> $SIG{__WARN__} = sub { print "<<<$_[0]>>>" }

  DB<121> warn "xxx" 
<<<xxx at (eval 58)[multi_perl5db.pl:644] line 2.
>>>

you can also run perlcode in a regex which checks $2 explicitly

edit

like in

  DB<133> $x =~ s/ (?{ print "\$2 undefined" unless $2 })/ /g
$2 undefined

And finally: please note that you are already using two eval levels with /ee (why?) so how to run check code should be obvious.

[Anonymous Monk]

Even outside the debugger, you can establish a warnings handler:

#!/usr/bin/perl

use warnings;
use strict;


my $sod='ip6 is used for...';
my $i=0;
my @baza_tek=qw( (?^u:ip(\d)) "internet_protocol-$1$2" );

{
    local $SIG{__WARN__} = sub {
        warn @_;
        warn "\$baza_tek[$i] is '$baza_tek[$i]";
        my $j = $i + 1;
        warn "\$baza_tek[$j] is '$baza_tek[$j]";
    };
    $sod=~s#$baza_tek[$i]#$baza_tek[$i+1]#gee;
}

And the reason $2 is undefined is that your regular expression has only one capture group, (\d).

[LanX]

?

> Even outside the debugger, you can establish a warnings handler:

Yes, since "for debugging" doesn't mean "inside debugger"

> And the reason $2 is undefined

see Re: On error, how to show the regexp values? (?: doesn't capture)

> has only one capture group, (\d).

capture groups are unrelated to \d

The outer parens would capture without the colon :

[nikolay]

Awesome! Thank you very much!

[nikolay]

But farther testing, on real data (that is much and is taken from different sources) did show, that the kind of exception, shows not the culprit at the time the error occurs. -- Like, if we have 1 000 patterns, and pattern #345 contains the error in focus, then the exception will show, say, pattern #378 -- back or forth in the row of patterns from the problem pattern, i do not know, having brought that exapmple just for better illustration of what i mean.

Personally, i think that the PERL mechanism for error handling, simply is late, and except it can be fixed, i think additional peace of code needs to be written like above have said -- that checks every pattern regexp, but, again, then we will get a lot matches, among whom will be the culprit.

So, before getting that mess of results, in order to make things done, i ask here for "wisdom" -- as it is kept saying here. ;-)