Remark: Suggestions on integrating smartypants?

[KyleAMathews]

Playing around with https://github.com/wooorm/retext-smartypants and Remark where I visit every text node and run it through smartypants. This works more-or-less but it'll miss converting quotes that are split into separate text nodes...

Thoughts?

[KyleAMathews]

Also it's not crazy performant as it's running retext a gazillion times.

[wooorm]

I unfortunately have no idea how to properly do that, with retext, currently.

It’s possible for the markdown tree to be converted to a natural language tree, through remark-retext. This keeps positional info intact, and strips the syntax. Thus making it possible for, say, [retext-spell][https://github.com/wooorm/retext-spell] to detect ex*em*ple as misspelt and suggest example instead.

That three can also be properly modified by retext-smartypants.

However, theres no way to merge that tree back into the markdown tree. And I’m not sure how to do it.

Say, staying with the earlier example, retext-spell autocorrected ex**em**ple to example. What happens with the strong node? There’s probably ways to deal with that. Prosemirror deals with similar problems, and things like React or virtual-dom of course. But I’m not entirely sure where to start.

😬

[wooorm]

Thought about this a short while, we probably need:

• [x] [unist-diff](https://github.com/syntax-tree/unist-diff) (Array.<Patch> = diff(Node, Node))
• [ ] unist-patch (patch(Node, Array.<Patch>))
• [ ] mdast-util-patch-from-nlcst (Array.<Patch> = patchFromNLCST(Node, Array.<Patch>))

the diff and patch would be ok, but changing nlcst patches to mdast patches is hard, I think.

[wooorm]

I’ve updated the above list as unist-diff is now a thing!

[KyleAMathews]

Nice! These look like super useful utilities (even beyond this use case).

[wooorm]

Still unsure how to do this. I never really got the utilities to work. Transforming stuff with retext and then merging the changes back into remark is, well, hard.
Something for a far future, maybe!