Much more conservative highlighting for 1.12

[tecosaur]

Following a few vigorous complaints about the highlighting scheme introduced by JuliaSyntaxHighlighting, I've pared back the default behaviour to something much more conservative. Now the only highlighting applied is:

• keywords (red)
• strings/char (green)
• comments (grey)

Old behaviour (1.11)

Nightly behaviour

Behaviour here

[christiangnrd]

I think this is an improvement over both 1.11 and current 1.12 behaviour.

[topolarity]

Is it possible to keep just the type-highlighting? Specifically in this part:

ncodeunits(s::AbstractString) -> Int

but not in the general code examples (i.e. only the documented signatures)

[tecosaur]

Is it possible to keep just the type-highlighting? Specifically in this part

Hmmm, it's possible but not particularly easily I'm afraid. I think we'd need to implement a new inferred code block "language" like julia-signature and then modify the way that highlighting is applied to it.

[JeffBezanson]

Is JuliaSyntaxHighlighting converting the :: for return type declaration to ->? I do prefer using the native syntax for this.

[mcabbott]

Note that both the nightly and PR examples above have weird colours for chars:

I'm not so sure that colouring strings & chars adds much.

It would be better to find a docstring which uses some keywords, and has some comments. Comments seem like the obvious first thing to treat specially. (Edit: I can't see how to try this out -- the PR does not contain the changes. But docstring for view has comments.)

[JeffBezanson]

Yeah that's not ok. If that can't be sorted out we'll have to drop coloring of chars and strings.

[christiangnrd]

Is JuliaSyntaxHighlighting converting the :: for return type declaration to ->? I do prefer using the native syntax for this.

That came from #57012

[tecosaur]

Note that both the nightly and PR examples above have weird colours for chars:

Urgh, when did that start happening? Good catch, I'll look into it…

[JeffBezanson]

Is JuliaSyntaxHighlighting converting the :: for return type declaration to ->? I do prefer using the native syntax for this.

That came from #57012

Ah. Should that be backported?