Links with complicated hashes aren't fully linkify-ed

[mattfeury]

I see this mostly with using Mixpanel, where we get links like:

https://mixpanel.com/report/1234/funnels/#view/1234/from_date:0,segfilter:(filters:!((filter:(operand:!(app,App),operator:%3D%3D),property:utm_medium,selected_property_type:string,type:string),(property:ab-test-01-08-14,selected_property_type:string,type:string)),op:and),segment_type:string,to_date:0

for sharing analytics tools. flowdock links it as:

https://mixpanel.com/report/1234/funnels/#view/1234/from_date:0,segfilter:(filters:!((filter:(operand:!(app,App),operator:%3D%3D),property:utm_medium,selected_property_type:string,type:string),(property:ab-test-01-08-14,selected_property_type:string,type:string)),op:and),segment_type:string,to_date:0

haven't looked at this code much. if someone can drop me some pointers to where this is in the code, I can try and take a look.

[phadej]

There url are matched with a regular expression, which is combined with a small helper "library":
the final one is on line 233

The match groups aren't really used for anything. The whole match is used only, but there are lot's of corner cases^† which makes the regexp quite complicated.

† for example:

• don't eat dot or comma after the link,
• or don't allow & before ?
• allow only matching pairs of parentheses in url's "path" part, so currently nested parentheses aren't allowed anyway

I added your case and few additional simpler cases, so you could give a try in fixing the regexp, without breaking other tests :)