undefined | Better HN

0 pointsgorhill9y ago0 comments

> don't use it to parse things that have rigorous standards describing them

Where a regex must be used, there is a reference regex for parsing URL: https://tools.ietf.org/html/rfc3986#appendix-B

Edit: a permalink to demonstrate the above reference regex: https://regex101.com/r/yJ5nU4/1 -- would have prevented the LastPass bug.

0 comments

sebcat9y ago

I once came across a bug in a codebase I inherited that parsed URI-references from HTML documents. That bug was caused by using that very same regular expression.

The string "foo" correctly gets parsed as a path (group 5). However the string "foo;key:value" gets parsed as a "foo;key" scheme (group 1,2) and a "bar" path (group 5) by the regular expression, but as a path if you follow the grammar.

scheme = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )

vs.

^(([^:\/?#]+):)?

:; are reserved characters, but they do not need to be encoded when they are a part of the path.

So yeah, don't use regular expressions for parsing.

j / k navigate · click thread line to collapse