Link parsing doesn't correctly match URLs when they include brackets

[grafana-dee]

For a given piece of Markdown:
[disambiguation](http://en.wikipedia.org/wiki/Disambiguation_(disambiguation))

I would expect to see the output:
<p><a href="http://en.wikipedia.org/wiki/Disambiguation_(disambiguation)">disambiguation</a></p>

However blackfriday moves the second ) outside of the link, thus breaking the URL:
<p><a href="http://en.wikipedia.org/wiki/Disambiguation_(disambiguation">disambiguation</a>)</p>

You can recreate this with this example code:

package main

import (
    "fmt"

    "github.com/russross/blackfriday"
)

func main() {
    md := []byte(`[disambiguation](http://en.wikipedia.org/wiki/Disambiguation_(disambiguation))`)

    html := blackfriday.MarkdownCommon(md)

    fmt.Println(`Output  : ` + string(html))
    fmt.Println(`Expected: <p><a href="http://en.wikipedia.org/wiki/Disambiguation_(disambiguation)">disambiguation</a></p>`)
}

The correct output is generated by most other Markdown parsers, here's Pandoc as the online link makes it the easiest to share:
http://johnmacfarlane.net/pandoc/try/?text=[disambiguation]%28http%3A%2F%2Fen.wikipedia.org%2Fwiki%2FDisambiguation_%28disambiguation%29%29&from=markdown&to=html

Bracketed links are very popular on Wikipedia and a few other top-100 internet sites.

[grafana-dee]

This also affects images:

![](http://www.broadgate.co.uk/Content/Upload/DetailImages/Cyclus700(1).jpg)

Becomes:

<p><img src="http://www.broadgate.co.uk/Content/Upload/DetailImages/Cyclus700%281" alt=""/> .jpg)</p>

When it should become:

<p><img src="http://www.broadgate.co.uk/Content/Upload/DetailImages/Cyclus700%281%29.jpg" alt=""/></p>

[rtfb]

This is definitely a bug, but a fix is a bit unclear. Skipping balanced pairs of parentheses is easy, but how far ahead should we look? Until EOL? Until a blank line?

[grafana-dee]

Gruber encountered this problem with the original Markdown library and helping to support others, he basically used a URL matching regexp that handled pairs of brackets in URLs which resolves the vast majority of instances in which it happens:
http://daringfireball.net/2010/07/improved_regex_for_matching_urls

\b((?:[a-z][\w-]+:(?:/{1,3}|[a-z0-9%])|www\d{0,3}[.]|[a-z0-9.\-]+[.][a-z]{2,4}/)(?:[^\s()<>]+|\(([^\s()<>]+|(\([^\s()<>]+\)))*\))+(?:\(([^\s()<>]+|(\([^\s()<>]+\)))*\)|[^\s`!()\[\]{};:'".,<>?«»“”‘’]))

Debuggex Demo

Note example 6 of the linked demo of that regex.

[rtfb]

Ooh, what a beastie :-)

I'll try to take a look if I can plug regexp in the middle of our other parsing and how will that affect the performance. Thanks!

[sdebruyn]

@miekg, can you do a PR to blackfriday with your fix?

[icco]

Yeah, I run into this almost daily still, right now I just hand replace parens with %28 and %29, but it's not a great fix.