p5pRT
commented
23 years ago From jfriedl@yahoo-inc.com
SUMMARY:
1) Gee\, it'd be nice to support (?\
2) I created a new magic variable $^N\, similar to $+.
3) If this is of interest\, is $^N a good name?
Feeling envy for highly advanced :-) languages like Java and Visual Basic\, whose regex languages allows named captures a'la
(?\
(sets $myval to the digits captured)\, I thought I'd use regex overloading to convert this syntaxt to
(\d+) (?{ $varname = $+ })
(capture to normal parens\, then use $+ to access it and assign to the variable)
This works fine except for a few details\, such as being able to have capturing parens nested within\, since $+ refers to the *hightest-numbered* set of parens used so far\, not the most-recently *closed* set of parens used so far.
So\, I created a new magic variable\, $^N\, that pretty much parallels $+ except it does indeed refer to the most-recently *closed* set of parens.
Now\, converting to
(\d+) (?{ $varname = $^N })
works even with nesting. This is very nice.
I can send the patches if wanted. But if wanted\, what is a good variable name? I picked $^N simply because I saw it was free.
For those interested\, I've appended my package to allow this\, and a short test program.
There are still "issues" with my overloading -- the variables named are not checked at runtime ('use strict' doesn't save you from referring to a nonexistant variable)\, and the variables are not "protected"\, so even if during a match a variable is set\, it won't be unset if the match later fails.
So\, it really would be nice if named captures were officially supported. I spent some hours digging through regcomp and regexec\, and succeeded only in killing massive amounts of neurons )-:\, so I don't think I'll be able to add it.
Jeffrey
Here is a short test script:
---snip------------------------------------------------------------------ use strict; use warnings; use Regex::SupportNamedCapture;
my $areacode; ## both lexical our $exchange; ## and global my $number; ## variables work fine.
"My number is 408-555-1212." =~ m{
\b
(?\
print "phone number is: ($areacode) $exchange-$number\n";
my $fullnumber;
"My number is 408-555-1212." =~ m{
\b
(?\
if ($fullnumber ne "$areacode-$exchange-$number") { print "Bummer\, you don't have \$^N support: fullnumber is [$fullnumber]\n"; } ---snip------------------------------------------------------------------
Here is the package:
---snip------------------------------------------------------------------ package Regex::SupportNamedCapture;
##
## This package allows regular expressions to have named captures\, a'la
##
## (?\
## ## Test to see if my proposed $^N is supported. ## Set $GutsResult to $^N if so\, set to $+ if not. ## my $GutsResult = do { no warnings; "1" =~ m/(1)/; if ($^N) { '$^N'; } else { '$+'; } };
our $OpenParens; ## needed for matching nested parens
my $NestedParenGuts = qr{
(?{ local $OpenParens = 0 }) ## counts the number of nested opens waiting to close
(?> (?> ## stuff not parens\, not escaped [^()\\]+
## escaped stuff | (?s: \\. )
# another opening paren | \( (?{ $OpenParens++ })
# a closing paren\, if we're expecting any | (?(?{ $OpenParens }) (?{ $OpenParens-- }) \)) )* ) }x;
##
## Mimics named capturing parens by converting something like
## (?\
$re =~ s{ (?\<! \\ ) # an unescaped... \(\? # "(?" \< # '\<' (\w+) # $1 - an identifier > # '>' ($NestedParenGuts) # $2 - regex guts \) # ')' }{ my $id = '$' . $1; my $guts = convert($2); "(?:($guts)(?{ $id=$GutsResult }))"; }exg;
return $re; ## return mangled regex }
1; ---snip------------------------------------------------------------------
Migrated from rt.perl.org#7190 (status was 'resolved')
Searchable as RT7190$