Re^3: Strange Occurrence in Substitution Statement

[roho]

Thank you for the possible solution Choroba! I tried it and it works. It looks like the non-capturing group prevents the last successfully matched regular expression from being used when the pattern is empty.

[LanX]

I think it's simpler, the pattern isn't empty, because (?:...) is always part of the pattern. Meta characters count too.

This idiom is meant to be used as s//something/ in complicated parsing, kind of inherited from sed/awk IIRC.

You should also be aware that a genuinely empty pattern IS always matching. In your case the replacement is just invisible because empty.

So I'd rather prefer to make this explicit and self documenting

• s/$extra// if $extra

  Nowadays I doubt there are more than a fraction of a percent of programmers knowing/using this feature.

  I'd really like to see a Perl version/feature to switch it off by default.

[choroba]

> I'd really like to see a Perl version/feature to switch it off by default.

I'd love that. And introducing something like /(*PREVIOUS)/ for those who need it.

[LanX]

> And introducing something like (*PREVIOUS) for those who need it.

Exists already since 5.37.10 as ${^LAST_SUCCESSFUL_PATTERN}

[roho]

Thanks LanX. I agree that this feature should be off by default. I think it falls into the category of being a little TOO helpful, with the side effects being very difficult to pin down. I spent days trying to isolate the cause and came up empty until posing the question here. Thanks to everyone for your help.