[Yaml] ParseException: pcre.backtrack_limit reached #22067
Conversation
| @@ -171,6 +171,7 @@ public function parse($value, $flags = 0) | |||
| $this->refs[$isRef] = end($data); | |||
| } | |||
| } elseif ( | |||
| // this regexp is backtracking-heavy and makes the parser fail on long strings | |||
| self::preg_match('#^(?P<key>'.Inline::REGEX_QUOTED_STRING.'|(?:![^\s]+\s+)?[^ \'"\[\{!].*?) *\:(\s+(?P<value>.+?))?\s*$#u', $this->currentLine, $values) | |||
There was a problem hiding this comment.
Most quantifiers in this regex could probably become possessive ones actually, to remove backtracking.
There was a problem hiding this comment.
Changing the regex to self::preg_match('#^(?P<key>'.Inline::REGEX_QUOTED_STRING.'|[^ \'"\[\{!].*?) *+\:(\s++(?P<value>.+?))?\s*+$#u', $this->currentLine, $values) doesn't seem to be enough.
Do you see another possible improvement?
ec69e9d to
f0256f1
Compare
|
Regexp fixed. PR ready. |
| @@ -171,7 +171,7 @@ public function parse($value, $flags = 0) | |||
| $this->refs[$isRef] = end($data); | |||
| } | |||
| } elseif ( | |||
| self::preg_match('#^(?P<key>'.Inline::REGEX_QUOTED_STRING.'|(?:![^\s]+\s+)?[^ \'"\[\{!].*?) *\:(\s+(?P<value>.+?))?\s*$#u', $this->currentLine, $values) | |||
| self::preg_match('#^(?P<key>'.Inline::REGEX_QUOTED_STRING.'|(?:![^\s]++\s++)?[^ \'"\[\{!].*?) *\:(\s++(?P<value>.+))?$#u', $this->currentLine, $values) | |||
There was a problem hiding this comment.
@GuilhemN the issue was the last part of the regexp, the related to the last \s* and the ungreedy quantifier just before.
| @@ -205,7 +205,7 @@ public function parse($value, $flags = 0) | |||
| $mergeNode = true; | |||
| $allowOverwrite = true; | |||
| if (isset($values['value']) && 0 === strpos($values['value'], '*')) { | |||
| $refName = substr($values['value'], 1); | |||
| $refName = substr(rtrim($values['value']), 1); | |||
There was a problem hiding this comment.
rtrim($values['value'], ' ')? tabs should not be trimmed.
There was a problem hiding this comment.
weren't they previously already, with this \s+$ tail?
There was a problem hiding this comment.
Indeed, it's legit then 👍
|
Thank you @nicolas-grekas. |
|
@nicolas-grekas @fabpot @stof this change breaks the parser when you a have a trailing space after the colon. Doctrine ORM test suite has it and it started to break after this PR got merged (we use I've discovered that by adding the following to the public function testCanParseContentWithTrailingSpaces()
{
$yaml = <<<YAML
items:
foo: bar
YAML;
$expected = [
'items' => ['foo' => 'bar'],
];
$this->assertEquals($expected, $this->parser->parse($yaml));
}I'm not sure if having a trailing space is valid according to the YAML syntax rules, however it was working before. |
|
Valid report, please open an issue. |
while merging 3.2 into master, I noticed that
testCanParseVeryLongValueis triggering this error on master, due to this regexp that we added for handling yaml tags. This regexp needs to be fixed so that we can merge the test case.ping @GuilhemN