And rightly so! Marked indeed is a fine š ā¬ parser!
How would you compare your approach to e.g. markdown-jsĀ¹, mdown-parse-pegjsĀ², or text.jsĀ³ (which are based on PEG parsing)?
I would like to elaborate a bit on the specific differences/benefits between the many MD parsers. (Iām maintaining an inventory of Markdown resources in a repo on Github.ā“)
Speed was originally marked's top priority. I would say marked's approach takes advantage of the fact that v8 generates _extremely_ fast machine code for regexes. This is just something I discovered through many endless nights of benchmarking the entire markdown test suite.
Using complex regexes for the lexemes seems like a stupid idea to most people, but it allowed me to optimize marked like crazy, so much so that it beat discount(written in c) in benchmark times (It might be a little bit slower now due to the latest features. I decided to ignore speed for a while to focus on features - will get back to optimization eventually).
The downside of all of this: marked loses some extensibility, however, we implement as much extensibility as possible by exposing the token stream and renderer.
Accuracy and sanity quickly became marked's next top priority. There's a certain threshold of accuracy you want in a markdown engine (you don't want it to be too accurate because markdown.pl had a lot of bizarre behavior). That's a different story.
I'll keep this as short as possible, but that's not easy.
Speed: It's relatively slow (but fortunately faster than markdown.js at the very least).
Quality: It carries with it a bunch of nonsensical quirks. Some of these quirks are inherited from the original markdown, some are unique to showdown. (I feel like I could write a treatise on list behavior alone, and how ridiculously showdown and markdown.pl handle them). It's not what people are accustomed to/find intuitive today. Here's a few problems (with showdown, as well as other engines) presented in a more tangible way: https://github.com/chjj/marked/blob/master/doc/broken.md
Code quality and maintenance: Showdown is a line-for-line port of Markdown.pl. As you can guess, this is not ideal, especially when you consider that the original markdown was written extremely awkwardly - it's essentially a pile of regexes doing global replaces on a single string to produce the output. This is _extremely_ confusing and messy to those who are not familiar with a code, and the original author of showdown, as far as I know, no longer maintains the project.
the world needs a new infrastructure for thoughtful documents,
including a new text-editor. but from the look of it thus far,
this github thing lacks _simplicity,_ by an order of magnitude.
How would you compare your approach to e.g. markdown-jsĀ¹, mdown-parse-pegjsĀ², or text.jsĀ³ (which are based on PEG parsing)?
I would like to elaborate a bit on the specific differences/benefits between the many MD parsers. (Iām maintaining an inventory of Markdown resources in a repo on Github.ā“)
[Ā¹] https://github.com/evilstreak/markdown-js [Ā²] https://github.com/shamansir/mdown-parse-pegjs [Ā³] https://github.com/sheremetyev/texts.js [ā“] https://github.com/rhythmus/markdown-resources