Remark stringify throws on list nested directly within a list

[erquhart]

Given the following html: <ul><ul><li>test</li></ul></ul>, remark-stringify errors out - the bullet param isn't defined when listItem is called. I'm assuming this means lists simply aren't expected to contain lists as direct descendants, but it would be nice to handle this gracefully, especially since rehype-remark is capable of generating the invalid mdast.

I have a live reproduction on Runkit, where you can also view the source mapped error:
https://runkit.com/erquhart/595544cb1cba570012811573

Also, here's a pretty-print of the MDAST that's causing the error:
https://gist.github.com/erquhart/72afed6fdf14fbf92f68d0f8ea163a88

I'm looking into it, but wondering if the problem is obvious to a maintainer.

[wooorm]

Hey @erquhart, thanks for the detailed report.
This is a problem because the tree is malformed (there’s a list directly in another list). That’s invalid, as there should always be a list-item between them. Thanks for finding it!

This is:

1. a bug in rehype-parse
2. not documented anywhere, and should probably be in mdast
3. maybe there should be a useful error thrown in listItem() if a non-list-item is given.

[erquhart]

It does seem to be documented in mdast, actually: https://github.com/syntax-tree/mdast#list

List (Parent) contains ListItems. No other nodes may occur in lists.

I'll close this and move it over to rehype-parse. Thanks!

[wooorm]

Yeah, I saw a note, but just made it a bit more explicit!

Don‘t worry, you can keep it open. It’s hard to keep track of all the repo’s, so I’ll just work on the others and close this from there!

[erquhart]

Okay, cool. Thanks :)

[wooorm]

I no longer think it’s a bug in rehype-parse, as it’s invalid HTML, the spec states the below on the <ul> element:

Content model:
Zero or more li and script-supporting elements.

...and the <ul> is neither <li> nor script-supporting.

What do you expect to happen on this invalid markup?

[erquhart]

It's tough because it works in browsers, and it occurs in popular tools like Paper from Dropbox. If you paste a nested list from Paper into an editor with Remark parsing, it breaks. I wouldn't always expect valid html to be pasted. HTML tidying isn't really something that can happen reliably in any of the js libs out there, either, not without a lot of downsides.

[erquhart]

Perhaps either in rehype-remark or remark-stringify, list item nodes can be added where needed?

[erquhart]

Maybe a plugin would work. Writing one seems simple enough, just need to check direct descendants of lists and wrap them. Would there be unexpected complexity, maybe related to positioning?

[wooorm]

Yeah I’m feeling this should be handled in hast-util-to-mdast, but I’m just not entirely sure how to handle it.

Should all non-list-items be wrapped in a list-item?

[erquhart]

That would be awesome, and I can't think of a downside - any affected nodes would otherwise break on stringify.