my $d = $s =~ s/[\s\S]*^$m *: (.*)$(?:[\s\S]*)/$1/rm;
The expression [\s\S] to express "match any character" cries out for comment. I assume it is used to avoid the . (dot) metacharacter when promoted by //s to "dot matches all" status.
This rubs me the wrong way. If dot (with //s) matches all, why not just use it that way? (All code examples that follow enable warnings and strictures. Also note that the //r substitution modifier is only avaliable with Perl versions 5.14+.)
This is arguably clearer, with only the tiny problem that it doesn't work! Why not?c:\@Work\Perl\monks>perl -wMstrict -le "my $s = qq{aaa : AAA\n} . qq{bbb : BBB\n} . qq{ccc : CCC\n} ; print qq{[[$s]]}; ;; my $m = 'bbb'; ;; my $t = $s =~ s/.*^$m *: (.*)$(?:.*)/$1/rsm; ;; print qq{[[$t]]}; " [[aaa : AAA bbb : BBB ccc : CCC ]] [[BBB ccc : CCC ]]
Consider the (.*) capture group. With dot matching anything, it greedily grabs everything to the end of the string. To achieve an overall match, the regex still has to match $ at the end of the string, which is easy, and (?:.*) "zero or more of anything" after the end of the string, also easy. So capture group 1 and $1 now contain everything to the end of the string, which is substituted back into the string.
But the intent of (.*) was only to capture everything up to the $ anchor before the first embedded newline (due to //m). How to restrain dot?
One way would be to use a *? "lazy" modifier for the normally greedy * match quantifier: dot will then match as little as necessary to get to the first $ anchor.
Now we're getting somewhere!c:\@Work\Perl\monks>perl -wMstrict -le "my $s = qq{aaa : AAA\n} . qq{bbb : BBB\n} . qq{ccc : CCC\n} ; print qq{[[$s]]}; ;; my $m = 'bbb'; ;; my $t = $s =~ s/.*^$m *: (.*?)$(?:.*)/$1/rsm; ;; print qq{[[$t]]}; " [[aaa : AAA bbb : BBB ccc : CCC ]] [[BBB]]
But one could argue that the intent of "anything except a newline" is more clearly expressed by [^\n] and "capture as much as possible to the first newline" is better as ([^\n]*) (remember that the code must be maintained, one must assume forever, so clear intent is important).
(In this version, the $ anchor is redundant, but does no harm and arguably serves to further clarify intent.)c:\@Work\Perl\monks>perl -wMstrict -le "my $s = qq{aaa : AAA\n} . qq{bbb : BBB\n} . qq{ccc : CCC\n} ; print qq{[[$s]]}; ;; my $m = 'bbb'; ;; my $t = $s =~ s/.*^$m *: ([^\n]*)$(?:.*)/$1/rsm; ;; print qq{[[$t]]}; " [[aaa : AAA bbb : BBB ccc : CCC ]] [[BBB]]
Lastly, an example in my own preferred style, taken from TheDamian's PBP:
The $m is no longer defined as a raw string, but with qr// as a regex in its own right. This allows it to be used "atomically" within another regex, as it is in the substitution: expressions like $m+ or $m{4} work as expected. The $ is still redundant, but still arguably clarifies intent. The same could be said about the preceding ^ in the regex, but I would argue that anchoring the $m atom in some way is potentially important, so just leave it be.c:\@Work\Perl\monks>perl -wMstrict -le "my $s = qq{aaa : AAA\n} . qq{bbb : BBB\n} . qq{ccc : CCC\n} ; print qq{[[$s]]}; ;; my $m = qr{ bbb }xms; ;; my $t = $s =~ s{ .* ^ $m [ ]* : [ ] ([^\n]*) $ .* }{$1}xmsr; ;; print qq{[[$t]]}; " [[aaa : AAA bbb : BBB ccc : CCC ]] [[BBB]]
And that's the first several inches of the whole nine regex yards. HTH
Give a man a fish: <%-(-(-(-<
In reply to Re^3: Why multiline regex doesn't work?
by AnomalousMonk
in thread Why multiline regex doesn't work?
by nbd
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |