@halirutan I looked at the highlight.js PR thread a couple of times, but it's a bit overwhelming, and I could not quite tell what the issues were. Therefore, I did not comment. As I understood, the disagreements were about finding appropriate semantic categories for each kind of highlighted thing. But I did not quite see why this was a big issue. Can't patterns just be param?

The thing is that no category is going to be 100% semantically correct because Mathematica is just very different, and the categories were made with other languages in mind (some of which also have very peculiar features, but their users may be too used to them to notice this)

Didn't comment as I felt I'd just add to the noise.

@halirutan That looks pretty good, I just wish operators were plain black.

@b3m2a1 It's "webby" and highlight.js was made by "webby" people. They use it, and see that it's different enough that the default categories don't apply, so they support it. It seems to be entirely subjective.

Is the problem the lack of willingness to allow "random" categories for special things like usages and In/Out, which are extremely useful to highlight differently? But they can't be highlighted differently unless using a non-semantically-correct category for them (because there just isn't a matching one)?

@halirutan @b3m2a1 When choosing categories, a very practical consideration is that highlighting should look reasonably good with the most commonly used themes. Maybe it's worth checking which high-profile websites use highlight.js and what is their default highlighting theme. I believe Discourse uses highlight.js

@Szabolcs The major point was that highlight.js tries to be general with "no language in mind", however, their CSS classes are missing even fundamental things like an "operator" style and on the other hand, there are specific styles for languages like JS, CSS and ReasonML.

That was my test highlighting where I checked if all parts are recognized correctly. I did another test with the old highlighter settings: