I don't like the idea of leftmost-outermost. That would severely slow the regex engine up if it has to decide whether it should store this capture or not each time it encounters one.

I'm also not that keen on assigning to an implicit hash or array.

That just means that again (as with $1, $2 etc.) I have to test for the value having been captured, and then re-assign it to where ever I want it before I call the next regex. The values captured from the first regex will either be over written (as would be the case if "the next regex" was actually the same one in a loop) or discarded if it is a different regex. The main reason for my wanting named regex captures is to avoid these two problems.

Assigning them directly to where I want them seems simpler and more reliable to me.

I don't like the idea of leftmost-outermost. That would severely slow the regex engine up if it has to decide whether it should store this capture or not each time it encounters one.

Presumably this would be resolved at the compile time of the regex, and IMO would be quite simple to resolve.

Assigning them directly to where I want them seems simpler and more reliable to me.

Hmm. I see your point (sort of) but I wonder how your angle would work with using qr// multiple times. For (not such a great) instance:

my $sep=qr/(?<sep>[-\/\\])/; 
my $date=qr/(?<year>\d{4})$sep(?<month>\d{2})$sep(?<day>\d{2})/
[download]

Hence the reason I thought of a hash of arrays. Also, the embedded $var name would _totally_ violate the quoting rules of perl, perhaps the idea that the names map to lexicals in scope (if defined) otherwise to ones it defines itself? (Which incidentally turns regex LHS's into a variant of a my(). I really dont see how this would work without seriously changing things. Mapping to a hash wouldn't have anywhere near the number of bizarre sideffects.



---
demerphq



First they ignore you, then they laugh at you, then they fight you, then you win.
-- Gandhi