Rinku should assume HTML encoded input

[rtomayko]

I'm seeing a lot of issues that seem to stem from Rinku trying to detect whether input is HTML encoded or not. #6 is one example.

Here's another:

>> Rinku.auto_link("This text is HTML encoded: http://example.com?x=1&y=2")
=> "This text is HTML encoded: <a href=\"http://example.com?x=1&amp;amp;y=2\">http://example.com?x=1&amp;amp;y=2</a>"

It's actually responsible for mangling the real text in this comment even. Here's the actual text I pasted:

https://gist.github.com/846f567ad158730811a2

Input into Rinku should always be HTML encoded. Why is it re-encoding in all these circumstances?

[vmg]

Hm. I remember having to re-encode it to comply with the Rails test suite... Let me look into that, I'm going to add a toggle.

[rtomayko]

I want to say the Rails autolinker tries to work with plain text (non HTML encoded) input, which is a fatal flaw in its design IMO. There's no way to correctly process this stuff unless you know whether the input is HTML encoded or not. I'd recommend mandating that input be HTML encoded.

With Rails 3 you could even check for html_safe? and fail fast or go ahead and do the conversion.

[vmg]

Your wish is my command. 1.4.0 removes all escaping, and performs the escaping on the Rails 3 monkeypatch method. How's that?

[rtomayko]

:metal:

[rtomayko]

Confirmed this fixes everything for us.