Color codes in docs are like chaos

[GuillaumeGomez]

A good example is: #37102 (comment)

Keywords should have one color code.
Enums should have another color code.
Structs should have another color code.
...

Color codes are up to debate. I'll start to clean this up soon.

cc @eddyb

[eddyb]

What kate does, which works well for me, is bold keywords (real bold, not like that ? in the linked issue),
and colors for things like macros or literals.
You can ignore one kind of information ("channel") to spot the other, easily. This principle can be extended to full HSL (each of hue, saturation and luminance being used for different semantic values).

If only one "channel" is used (e.g. hue), then it's much harder to distinguish between various intents.
That and striking colors should not be used for every other word, it leads to semantic satiation.

Example from kate, with red ? added locally (I really wish bold ? were bolder 😞):

[eddyb]

cc @rust-lang/docs (also not sure how to ping the style team)

[GuillaumeGomez]

It seems to consider Ok as keyword. :-/

[eddyb]

Well, it seems a bit silly:

        <list name="constants">
                <item> true </item>
                <item> false </item>
                <item> Some </item>
                <item> None </item>
                <item> Ok </item>
                <item> Err </item>
                <item> Success </item>
                <item> Failure </item>
                <item> Cons </item>
                <item> Nil </item>
        </list>

Out of those variants only Some, None, Ok and Err remain, and they're not keywords, unlike true and false. It's not exactly been thoroughly updated. However:

<keyword String="constants" attribute="Constant" context="#stay"/>

<itemData name="Constant"     defStyleNum="dsConstant" spellChecking="0"/>

And dsConstant is distinct from dsKeyword, so I can change it in the editor settings, if I want to.

[GuillaumeGomez]

You should! :p

[Mark-Simulacrum]

@GuillaumeGomez Is this something you still want to pursue? I've marked as T-doc.

[GuillaumeGomez]

@Mark-Simulacrum: It's still in discussion actually, so yes. :)

[steveklabnik]

Closing, as this is not super-actionable as is and many fixes have landed.