Re: regex to capture an unsigned decimal value, but also preserving the user's formatting.

[davido]

You can make the dot possessive and the trailing digits greedy but optional:

while(<DATA>){
    chomp;
    next unless length;
    say "$_ => [", (m/(\d*\.?+\d*)/x ? "$1]" : ']');
}

__DATA__
0.12
.12
12.
12

This yields the following output:

0.12 => [0.12]
.12 => [.12]
12. => [12.]
12 => [12]

I do not know what additional edge cases you might encounter that would break this. But for your test cases it works.

Two tricks here. First, it was a mistake to use the + quantifier on the trailing \d, because then it became a required portion of the match. If it's required, then 12. is going to fail to match the decimal point because it comes before the trailing \d+. By making the quantifier * we stay greedy, but optional.

The next trick is the ?+ quantifier for the decimal point. The + here signifies to be possessive -- once it has matched, hold onto what it matched against, and don't give it up during backtracking.

For reference, here are the two regexes (yours and mine) in close proximity. You can disregard the parens here; I'm only using them for capturing, which you were achieving by referring to the match hash.

m/( \d*  \.?+  \d* )/x    # Mine
m/( \d*  \.?   \d+ )/x    # Yours

I suggest walking through your original regular expression, and the one I've provided using the Regexp::Debugger's rxrx utility. I think my inadequate description will be clearer once you see the wheels in motion.

Update: I've fixed the backslashing of the . in the pattern.

[Athanasius]

Hello davido,

Your use of the + (possessive) quantifier is ingenious, as it leads to simple, elegant code. I think you need to backslash the dot to avoid matching any character:

say "$_ => [", (m/(\d*\.?+\d*)/x ? "$1]" : ']');
#                     ^

But my main quibble is that this doesn’t work consistently if the decimal is embedded in a longer string. (Whether that’s actually a requirement isn’t clear from the OP.) For that, I came up with the following regex, which is verbose but seems to work OK:

#! perl
use strict;
use warnings;
use feature qw( say );

my $decimal = qr{
                    (
                        (?: \d+ \.? \d* ) |
                        (?: \d* \.? \d+ )
                    )
                }x;

while (<DATA>)
{
    chomp;
    next unless length;
    say "$_ => [", (/$decimal/x ? "$1]" : ']');
}

__DATA__
0.12
.12
12.
12
no numbers here
abc.def42
.7zx

Output:

16:39 >perl 1621_SoPW.pl
0.12 => [0.12]
.12 => [.12]
12. => [12.]
12 => [12]
no numbers here => []
abc.def42 => [42]
.7zx => [.7]

16:39 >

Cheers,

[davido]

Correct on both counts: (1) The dot should have been backslashed. (Edit: I've fixed the backslashing of .) (2) This works only for strings that don't contain dots anywhere to the left of the target "number". :). I didn't see any in his sample input, but the possibility supports the notion of "You have a problem. You think 'I will solve it with a regular expression.' Now you have two problems."

A better solution would probably leverage Regexp::Common, or the regular expressions revealed in perlfaq4 under "How do I determine whether a scalar is a number/whole/integer/float?"

[AnomalousMonk]

I think the other important point to make is that the order of the sub-patterns in the ordered alternation is critical:

c:\@Work\Perl\monks>perl -wMstrict -le
"my $ufp = qr{ \d* [.]? \d+ | \d+ [.]? \d* }xms;
 ;;
 for my $s (qw(0.12  .34  56.  78)) {
   printf qq{'$s' -> };
   printf qq{'$1' \n} if $s =~ m{ ($ufp) }xms;
   }
 ;;
 my $s = '0.12 -0.98 .34 -.76 56. -54. 78 -32 bla bla abc.def42 .7zx';
 printf qq{'$1' } while $s =~ m{ ($ufp) }xmsg;
"
'0.12' -> '0.12'
'.34' -> '.34'
'56.' -> '56'
'78' -> '78'
'0.12' '0.98' '.34' '.76' '56' '54' '78' '32' '42' '.7'

(See  '56.' and  '-54.' instances.)