Bug #79257 [Ver]: Duplicate named groups (?J) prefer last alternative even if not matched

Edit report at https://bugs.php.net/bug.php?id=79257&edit=1 ID: 79257 User updated by: ptomulik at meil dot pw dot edu dot pl Reported by: ptomulik at meil dot pw dot edu dot pl -Summary: Duplicate named groups ((?J) modifier) broken since 7.4 +Summary: Duplicate named groups (?J) prefer last alternative even if not matched Status: Verified Type: Bug Package: PCRE related Operating System: Debian/Ubuntu -PHP Version: 7.4.2 +PHP Version: 7.3+ Block user comment: N Private report: N New Comment: This actually happens on all versions, if we consider the following example <?php preg_match('/(?J)(?:(?<g>foo)|(?<g>bar))(geez)/', 'foogeez', $matches, PREG_UNMATCHED_AS_NULL); var_dump($matches); ?> Expected result: ---------------- array(5) { [0]=> string(7) "foogeez" ["g"]=> string(3) "foo" [1]=> string(3) "foo" [2]=> NULL [3]=> string(4) "geez" } Actual result: -------------- array(5) { [0]=> string(7) "foogeez" ["g"]=> string(3) "foo" [1]=> string(3) "foo" [2]=> NULL [3]=> string(4) "geez" } Previous Comments: ------------------------------------------------------------------------ [2020-02-11 16:13:52] ptomulik at meil dot pw dot edu dot pl Looks like the actual problem existed in 7.3 and earlier, but it required slight modification to be reproduced. I'll try to update the bug report to take it into account. ------------------------------------------------------------------------ [2020-02-11 15:41:26] nikic@php.net Parts of this actually also seem broken on earlier versions, for example: preg_match('/(?J)(?:(?<g>foo)|(?<g>bar))(?<h>baz)/', 'foobaz', $matches); var_dump($matches); array(6) { [0]=> string(6) "foobaz" ["g"]=> string(0) "" [1]=> string(3) "foo" [2]=> string(0) "" ["h"]=> string(3) "baz" [3]=> string(3) "baz" } Key "g" should be "foo" here though. ------------------------------------------------------------------------ [2020-02-11 13:11:21] ptomulik at meil dot pw dot edu dot pl Description: ------------ It appears, that the "backward incompatible changes" in PHP7.4 - "Regular Expressions", as described here: https://www.php.net/manual/en/migration74.incompatible.php affect named capture groups when duplicate names are enabled ((?J) modifier/PCRE2_DUPMANES config). In 7.4 named capture group always returns the content captured in last alternative, no matter which of the alternatives matched. This differs from 7.3 and renders DUPNAMES feature useless. Test script: --------------- <?php preg_match('/(?J)(?:(?<g>foo)|(?<g>bar))/', 'foo', $matches, PREG_UNMATCHED_AS_NULL); var_dump($matches); ?> Expected result: ---------------- array(4) { [0]=> string(3) "foo" ["g"]=> string(3) "foo" [1]=> string(3) "foo" [2]=> NULL } Actual result: -------------- array(4) { [0]=> string(3) "foo" ["g"]=> NULL [1]=> string(3) "foo" [2]=> NULL } ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=79257&edit=1