Swizec Teller - a geek with a hatswizec.com

    How to debug unified, rehype, or remark and fix bugs in markdown processing

    Ever had to debug a compiler {{ subscriber.first_name | truncatewords: 1, "" | capitalize }}? It's hard.

    Typical engineering wisdom goes that if you blame the compiler, you're wrong. You messed up and can't figure it out and it is not the compiler's fault.

    But sometimes it is.

    Click through for source
    Click through for source

    I'm working on a script to convert Wordpress exports to Markdown and after months of tinkering, noticed a terrible bug. JSX code samples don't work.

    You take this bit of Wordpress source code:

    Click through for source
    Click through for source

    And you get this output:

    Click through for source
    Click through for source

    But you're supposed to get this:

    Click through for source
    Click through for source

    The compiler that translates plain Wordpress HTML to Markdown makes a mistake. Sees HTML inside a code sample, parses it as HTML, then silently drops it.


    How would you fix this?

    I verified the input is all correct until it goes into the html-to-markdown compiler. That's always a good first step. Garbage in, garbage out

    How compilers work


    Compilers work in 3 stages.

    1. The front-end parses text inputs into an Abstract Syntax Tree – AST
    2. The middle-end manipulates the AST to add optimizations and other changes
    3. The back-end turns your AST back to text

    That's the general idea. You can read about compilers in more detail in my LOLCODE-to-JavaScript compiler article.

    Modern tools for Markdown processing

    Your best bet to play with markdown right now is the Unified ecosystem. It's what Gatsby, Next, Prettier, and many others use.

    Unified is a system of plugins with plugins that let you parse, process, and output ASTs. Easy to get started and confusing to dig deep. Hard to find the code that does anything.

    Every plugin calls another plugin. πŸ˜…

    General lay of the land is that:

    • unified is the protocol that holds it all together
    • rehype is the plugin ecosystem for HTML parsing/rendering
    • remark is the plugin ecosystem for Markdown parsing/rendering

    A basic example that parses HTML and spits out Markdown goes like this:

    Click through for source
    Click through for source

    How you debug a compiler

    Debugging a compiler is where life gets tricky. A whole processing pipeline, oof.

    You're trying to find out which step of the processing pipeline makes a mistake. And whether it's a mistake at all. You could still be the one holding it wrong.

    Works for everyone else. Chances are you're holding it wrong. But how do you figure out how you're holding it wrong?

    You can watch me work through the issue here:

    Click through for source
    Click through for source

    I go through the process methodically injecting console.logs into the pipeline to see how nodes change shape. Trying to figure out where it goes wrong.

    In the end I decided it wasn't a bug. Or if it was, it's too deep for me to fix.

    A debugging kludge

    First you create a small example. As small as possible.

    Then you build a debugging kludge – a tiny plugin that helps you hook into the compilation process.

    It's a kludge because it's temporary, dirty, and for your eyes only. Like slapping a console.log into your code to see what's going on.

    Your pipeline has two sides, if you're translating HTML to Markdown:

    1. The Rehype side
    2. The Remark side

    These require different kludges because the AST structure is different. Tag naming is different, properties are different, etc. Both are tree structures.

    Debug a Rehype AST

    To see how Rehype sees your code blocks, you use a kludge like this:

    Click through for source
    Click through for source

    Look at the output and you see what Rehype sees. Every property, node structure, and value.

    Change the if to observe different nodes. I don't recommend printing the whole AST. Too much.

    Debug a Remark AST

    On the other side of the rehype2remark divide, you have a Markdown AST. Same basic structure, different naming conventions.

    Remark is easier to debug because it's well-supported by the unist ecosystem of helper plugins. πŸŽ‰

    To print code blocks on the Remark side:

    Click through for source
    Click through for source

    The visit helper handles recursion for you.

    How to fix a compiler pipeline issue

    There's 3 ways to fix a compiler pipeline issue:

    1. On the parsing side
    2. In the middle
    3. On the stringify side

    Fixing parsers by hand is hard. Writing, too. Most parsers are auto-generated from grammar definition files.

    Rehype parses my JSX-containing code blocks as if they were HTML. I thought they should be strings.

    After digging, I discovered that Rehype uses Parse5 – an HTML5 compliant parser used in the entire Node ecosystem. With 75,000,000 downloads per month ... yeah I'm the one who's wrong.

    im_wrong giphy

    Fixing stringification is boring. You'll dig and dig and do nothing fun for days. Shit's boring, we did it for my compilers class in college.

    Plus you've got a battle-tested stringifier right there. You need to give it the right input tree πŸ˜‰

    MacGyver a fix in the middle

    macgyver giphy

    Compiler middleware is the best place to add your fix.

    Take the AST, change it so stringification does what you want. Magic.

    Here's what I did for my example of JSX-containing code blocks:

    Click through for source
    Click through for source

    Find all the code blocks, replace their children with a fake new child.

    Click through for source
    Click through for source

    Make the child a <code> block because Remark likes that better and stringify the whole <pre> tag into a single text node.

    You get:

    Click through for source
    Click through for source

    Not perfect, but better. Need to fix the child html nodes or find a JSX stringifier πŸ€”


    Did you enjoy this article?

    Published on August 17th, 2020 in Uncategorized

    Learned something new?
    Want to become an expert?

    Here's how it works πŸ‘‡

    Leave your email and I'll send you thoughtfully written emails every week about React, JavaScript, and your career. Lessons learned over 20 years in the industry working with companies ranging from tiny startups to Fortune5 behemoths.

    Join Swizec's Newsletter

    And get thoughtful letters πŸ’Œ on mindsets, tactics, and technical skills for your career. Real lessons from building production software. No bullshit.

    "Man, love your simple writing! Yours is the only newsletter I open and only blog that I give a fuck to read & scroll till the end. And wow always take away lessons with me. Inspiring! And very relatable. πŸ‘Œ"

    ~ Ashish Kumar

    Join over 14,000 engineers just like you already improving their careers with my letters, workshops, courses, and talks. ✌️

    Have a burning question that you think I can answer?Β I don't have all of the answers, but I have some! Hit me up on twitter or book a 30min ama for in-depth help.

    Ready to Stop copy pasting D3 examples and create data visualizations of your own? Β Learn how to build scalable dataviz components your whole team can understand with React for Data Visualization

    Curious about Serverless and the modern backend? Check out Serverless Handbook, modern backend for the frontend engineer.

    Ready to learn how it all fits together and build a modern webapp from scratch? Learn how to launch a webapp and make your first πŸ’° on the side with ServerlessReact.Dev

    Want to brush up on your modern JavaScript syntax?Β Check out my interactive cheatsheet: es6cheatsheet.com

    By the way, just in case no one has told you it yet today: I love and appreciate you for who you are ❀️

    Created bySwizecwith ❀️