Add support for more selectors

[cstrahan]

Hello! As part of my work on Happybara, I've implemented a CSS selector parser and a converter from CSS to XPath:

https://github.com/cstrahan/happybara/blob/to-query/happybara/src/Happybara/XPath.hs#L40 https://github.com/cstrahan/happybara/blob/to-query/happybara/src/Happybara/CSS.hs

If you're interested, I'd like to fold my implementation into HandsomeSoup.

If you wouldn't mind taking a dependency on hxt-xpath, I think we should be able to use my css-to-xpath conversion directly, using getXPathTreesInDoc.

Thoughts?

[egonSchiele]

Give me a little time to read over the code, but this sounds like a great idea. As an aside, happybara sounds cool.

[cstrahan]

Thanks!

In the next day or two, I'm going to write some fuzzing tests. The plan for the tests:

• Create set of css selectors and the html documents to apply them to.
• Use Capybara to generate the expected nodes; dump that to json.
• In the Haskell test suite, load up the expectations, generate the xpaths, and compare results (using hxt-xpath).

So, while there aren't any tests now, there will be pretty soon :).

Another idea: if we merge my implementation (or parts thereof), it might make sense to emit an Expr instead of a string. That would be ideal for two reasons:

• Skip the cost of having hxt-xpath parse the generated xpath (and the string construction itself).
• Remove one more point of failure (e.g. malformed xpaths, or potentials bugs in hxt-xpath's xpath parser).

Users could then render the Expr with formatXPathTree, if they wanted to (or we could provide a function for convenience).

I'm going to wrap up the test suite, and then I'll hack on a PR for you to review.

[egonSchiele]

Hmm, did anything happen with this? Sorry it took me so long to get to, and I don't see any changes anymore.

[cstrahan]

@egonSchiele No worries; I got caught up in other things. I'd still like to take a stab at this, but it will probably be a couple weeks before I can dive back in.

[egonSchiele]

It's fine...I have a lot of other stuff on my plate right now anyway :)