Make the tokenizer able to be run separately

[aredridel]

Make the tokenizer able to be run separately

[gwicke]

I am very much interested in this as well. I am using the tree builder with a custom (MediaWiki) tokenizer in a prototype I am working on currently. For now I modified the parser to take a tokenizer as an argument in parse(), but this is just a quick hack that ignores tokenizer modes completely.

Right now I am mostly working on the tokenizer, but might be able to put some effort into the interface later.

[aredridel]

Oh, sweet. I'm integrating the latest draft's tokenizer changes, which flattens the modes out into separate states. That should help move this along.

[papandreou]

@aredridel: Has that work been completed? I'm thinking about updating a project to the newest version of html5 and getting rid of the state transitions in my code.

[aredridel]

It has not, sadly! It's not trivial to get 100% accurate, since the HTML5 algorithms assume a document tree, and if you're not parsing fully, you don't get one. I need to rework for some of the latest spec changes, and that should simplify, since they've flattened some of the parser down into tokenizer states.

[dgreensp]

I'm using just the tokenizer at the moment. I was able to pull it out, but it would be nice if it was explicitly separable.

[aredridel]

It's hard with the revision of the spec I was originally targeting, since the parser state feeds back into the tokenizer. With the latest revision, much if not all of this is flattened out. As I migrate toward the current parsing spec, it'll smooth that out.

[dgreensp]

Sounds great. Thanks for writing this package, btw, it seems to be one of a kind in the JS world.

[aredridel]

I'm kinda surprised at how one of a kind it is, but it's needed!