I got it to work by replacing the regexp with
I found that if I only changed $+ to $&, I did not get the correct output:qr/(?:$QUOTED|$NUM)(??{'' eq $& || '.' eq $& ? FAIL : SUCCEED})/ __END__ $VAR1 = [ 'name => ', '"foo"', ', fav.num => ', '3' ];
$VAR1 = [ 'name => ', '"foo"', ', fav', '', '.num => ', '3' ];
Update: Another way to fix the problem is to retain the test on $+ but make sure that the capture parens happen before the test (not around the test):
Note that we still have to check for an empty match. Also note that in this version there are no parens around $VALUE in the split.my $VALUE = do { use re 'eval'; qr/($QUOTED|$NUM)(??{'' eq $+ || '.' eq $+ ? FAIL : SUCCEED})/; }; my $text = 'name => "foo", fav.num => 3'; my @text = split /$VALUE/ => $text; __END__ $VAR1 = [ 'name => ', '"foo"', ', fav.num => ', '3' ];
The empty match (that necessitates the extra check in the ...?...:... expression) is caused by something in $NUM, though I have not yet pinned it down. The following gives a glimpse of what the engine is up to:
An empty match happens after each non-empty candidate fails.my $VALUE = do { use re 'eval'; qr/$NUM(?{ print "($`)($&)($')\n" })(?!)/; }; my $text = 'name => 5, fav.num => 3'; $text =~ /$VALUE/; __END__ (name => )(5)(, fav.num => 3) (name => )()(5, fav.num => 3) (name => 5, fav)(.)(num => 3) (name => 5, fav)()(.num => 3) (name => 5, fav.num => )(3)() (name => 5, fav.num => )()(3)
Update 2: Fixed bug pointed out by Roy Johnson. Thanks.
the lowliest monk
In reply to Re: Forcing a regex to fail
by tlm
in thread Forcing a regex to fail
by Ovid
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |