How to use remark-react with hightlight.js about remark-react

[wooorm]

I actually have a hacky but working AST-based re-implementation of highlight.js working now. It’s probably bug filled though, will have to add lots of tests (I believe there’s like 60 supported languages? Gah.) But do expect more to come soon! 😄

@tmcw I’m writing it to expose HAST nodes, a unist and minimal HTML syntax tree. This means to get this working with react there needs to be a transformation between the two. I don’t expect it to be very hard, but I’m not experienced with react. could you help me there when the time comes?

[tmcw]

Awesome, been a bit quiet on this because of work-work but love the way this is shaping up.

[bebraw]

I published the solution as remark-react-lowlight. You can see it in action at react-component-boilerplate.

It's probably safe to close this issue now.

[rgbkrk]

@bebraw is there a way to use that component without having to specify the highlight js lang (relying on code fence blocks)? The example in the README makes me think I'd have to specify it.

[tmcw]

It's not currently possible to use highlight.js: highlight.js produces a string of HTML content, and remark-react is expressly written to avoid such HTML escapes. If there was a highlighter that output an AST instead, then we could integrate it into remark-react - otherwise, highlight.js is not compatible with remark-react's security model.

[davegomez]

Do you have an example of the AST structure you would need to implement the highlighting?

[wooorm]

I was thinking about the same problem recently, for example, I’d like to support syntax highlighting in remark-vdom, remark-man too.

I did some digging into highlight.js, and I don’t think their per-language syntax files needs changing (for example, javascript, wouldn’t need changing). However, their core is tightly coupled with outputting HTML.

So, either someone needs to create a fork which supports other syntaxes (in the form of specifying a compiler), or highlight.js itself should be rewritten, but I’m not sure whether they’d want to.

I have a full plate, but I’ll try my hands on whether it’s feasible, but I’d suggest you’d raise an issue regarding decoupling on highlight.js’s repository.

[tmcw]

Sure! Let me know when you get to the point of integration

[wooorm]

Great! 👌

[wooorm]

Done! https://github.com/wooorm/lowlight

[bebraw]

FYI, here's a Node friendly highlighter that has been developed based on Prism: Illuminate.js. Prism deals with JSX better than Highlight.js.

It would be great if you could design the system so that it's possible to glue in Illuminate. Note that there's react-illuminate as well.

[wooorm]

@bebraw just saw your comment, that looks nice too, although it doesn’t support many language, and it seems to be in beta?

@tmcw I think all the building blocks for syntax highlighting with remark-react are now here, there’s lowlight (or others), and there’s react-syntax-highlighter (which uses lowlight under the hood).

What remains is “component” support in remark-react, which I guess means that users of this remark plugin can specify that code is rendered as SyntaxHighlighter?

Again, I’m not very familiar with React though, so you probably have a better idea of how this should be handled :)

[bebraw]

@wooorm It's a little limited at the moment, that's true. Given it's Prism based, it would be possible to add support to the missing languages if needed I think. I opened an issue so we get feedback from the author. Something Prism based would be a winner solution for me given how well it highlights.

[wooorm]

Diversity is good! If you want prism, that’s fine and it should happen.

I think this issue will allow that, if we add support for components through something similar like the method I proposed above!

[bebraw]

It would be nice to get this moving again. Can someone in the know explain what should happen code-wise? I could try to look into making the changes.

[wooorm]

I don’t know the API of react, but you can use react-lowlight or react-syntax-highlighter, both based on lowlight, to do virtual (thus, safe), syntax highlighting. Then, wrap one of those components, and integrate with remarkReactComponents (see the readme of remark-react). I’m not entirely sure what API they use, and if you can access the lang of a code node though.

@tmcw, could you clarify whether this should work?

[bebraw]

@wooorm Thanks. Esp. react-lowlight seems great. I think could be enough if I just replace the portion that does code now with that component. I'll give it a quick go and report back.

[bebraw]

I gave it more thought. We probably should agree on a basic interface. Right now we have

remark().use(reactRenderer)

What if we did

remark().use(reactRenderer({
  code: codeRenderer // custom logic would go here
}))

The idea is that this type of hook system would allow you to replace h per node type. This would allow the maximum amount of customization without breaking backwards compatibility. Most importantly it would enable highlighting plugin.

[wooorm]

I believe the current code already supports something like that:

var remark = require('remark');
var toReact = require('remark-react');

remark().use(toReact, {
  remarkReactComponents: {
    'code': 'Lowlight'
  }
}))

Then, Lowlight is passed to React.createElement. But, aside from supporting string, maybe function should be supported in remark-react too? Or, create a wrapper component around Lowlight which supports the attributes given by remark-react and transforms them to support the attributes expected by react-lowlight.

[bebraw]

Oh, yeah. Missed that. It would need a tiny adapter right there. 👍

[wooorm]

Maybe create remark-react-highlight? So people who use remark, react, and remark-react, need just remark().use(remarkReact).use(remarkReactHighlight)? That’d be my suggestion :)

[bebraw]

I managed to set up highlighting to work with remark-react. Here's a quick demo.

@wooorm How would that remark().use(remarkReact).use(remarkReactHighlight) work? I would need to understand how to connect the plugins internally to patch remark-react behavior (literally that element definition). Alternatively we could do something like this:

{remark().use(reactRenderer, {
  remarkReactComponents: {
    code: RemarkLowlight({
      // list supported languages here, lowlight needs them
      js: jsLangugeDefinition
    })
  }
}).process(readme)}

Lowlight requires those language definitions. Instead of passing them explicitly we could hide the registration within the package itself and perhaps support optional registration that would override that default. I think the reasoning is that each language definition bloats the bundle so the decided to go with opt-in kind of system there.

[wooorm]

Oh, you’re right, it wouldn’t work (currently) as a remark plug-in :/ I like your proposed idea; supporting an object mapping names to language definitions.

And, indeed, that’s the reason. Lowlight suggests against subscribing all languages when in the browser.

[wooorm]

Nice! Maybe add a link to the README here though?

[rgbkrk]

Ah, I see. If it's in the object passed to RemarkLowlight({}) it gets registered. Noticed that it will break down if the language specified in a code fence block isn't registered though.

[bebraw]

@rgbkrk Yeah. It has to register the languages first. Can you open an issue about the breakage at my repo? Maybe the right way to solve that is to skip highlighting in that case and just give a warning instead.

[mattcreager]

Hey I'm attempting to get remark-react and syntax highlighting to cooperate, hoping someone who's run the gauntlet before can help :)

My naive implementation looks something like

Lowlight.registerLanguage('js', js);

const Pre = ({children}) => {
  return <Lowlight language={'js'} value={children[0].props.children[0]} />
}

Pre.propTypes = {
  className: React.PropTypes.string,
  children: React.PropTypes.node,
};
...
const remarkReactComponents = { pre: Pre };

Which would work if I only needed to highlight js,. so from here I wrote a little parser that grabs the language from a node and drops it into an attributes object.

node.attributes.className = ['hljs', `language-${lang}`];

Which never appears to be available on the props of my React nodes.

P.S. I'm using pre rather than code to prevent nested elements.

[wooorm]

I think @bebraw got something working with https://github.com/bebraw/remark-react-lowlight.

@bebraw, does remark-react-lowlight support pre?

[bebraw]

@bebraw, does remark-react-lowlight support pre?

I can't see anything pre specific at the code. Maybe "it just works"™. Worth a go.

[wooorm]

Cool, I’ll whip up a PR to include this in the docs here!