URI.withinString does not parse URIs that contain balanced parentheses

[astorije]

These 2 examples are correctly parsed by GitHub so this is a good place to demonstrate:

1. https://example.com/with_(balanced_parentheses)
2. https://example.com/with_unbalanced_parenthesis)

URI.withinString will get case 1 correctly, but not case 2.

Consider this real-life example: http://en.wikipedia.org/wiki/The_Martian_(Andy_Weir), here is what happens using the latest version on npm ([email protected]):

> URI.withinString('https://en.wikipedia.org/wiki/The_Martian_(Weir_novel)', function (url) { return '<a>' + url + '</a>'; });
'<a>https://en.wikipedia.org/wiki/The_Martian_(Weir_novel</a>)'

The second parenthesis gets pushed out.

Other bracket types (`[] {} <>') should follow the same rule.

[rodneyrehm]

Nice catch! The problem looks simple enough to allow someone else to try fixing it. Here are some pointers to help with that:

URI.withinString uses the URI.findUri expressions to hone in on potential URLs. The expression URI.findUri.trim is used to trim those characters, because URI.findUri.end is only looking for some whitespace.

I guess I would try splitting URI.findUri.trim into opening and closing characters and only trim closing chararcters, if the identifed URI does not contain the corresponding opening character. Adding a test for the change would likely help, too.

[astorije]

Hi @rodneyrehm, thanks for the pointers! I'll see if I can send a PR within the next few days and/or come back with questions. But yes, my tests were already written when your answer came in :-)

[blaenk]

Any news on this?

[davericher]

bump

[rodneyrehm]

since there was no patch provided by the community, I patched it myself…

The characters (), [], {} and <> are considered and the closing character is included in the URL if it also contains the opening character:

http://example.com/foo(bar)
-> "http://example.com/foo(bar)"

http://example.com/foo(bar))
-> "http://example.com/foo(bar)"

http://example.com/foobar))
-> "http://example.com/foobar"

released as v1.18.4.

[astorije]

Thanks a lot, @rodneyrehm! Sorry I wasn't helpful on that.