Improved tree-sitter queries

[kirawi]

@variable.property -> @variable.other.member
@property -> @variable.other.member
@number -> @constant.numeric.{integer, float}
@escape -> @constant.character.escape
@constant.character is now uniformly used for all chars.

[kirawi]

Should this be pulled out to match all tuple_patterns? I don't know if there are any other cases where it's used that we probably don't want to highlight it as a variable. I'll do it later once I refactor highlights.scm to completely cover all highlighting based on tree-sitter-rust/grammar.js

[archseer]

Not sure, we could do that and simplify

[archseer]

variable.field: What you really want is variable.other.member as declared by Sublime Text:

Fields, properties, members and attributes of a class or other data structure should use:

Here's an example of atom's tree-sitter Rust grammar using it: https://github.com/atom/atom/blob/fd5577910528774b549e94407b555da89f55db79/packages/language-rust-bundled/grammars/tree-sitter-rust.cson#L39

[kirawi]

I feel that people wouldn't understand what that means just by looking at it, versus variable.field.

[archseer]

A struct member is pretty common terminology. I'd rather stay in line with TextMate/Sublime scopes since that's what most grammars already use (including VSCode)

[archseer]

I'm also looking at other grammars to see if type.enum.variant is a good choice or not for enums, doesn't look like Atom styles it. VSCode probably has a more detailed grammar

[kirawi]

Looks like variable.other.enumMember

[archseer]

The declaration is type.enum, can't figure out the variants inside code though. https://github.com/microsoft/vscode/blob/e78e63fc3af3113d0a32cb479b8e7c232901d4e4/extensions/rust/syntaxes/rust.tmLanguage.json#L995

Result and Option are special cased https://github.com/microsoft/vscode/blob/e78e63fc3af3113d0a32cb479b8e7c232901d4e4/extensions/rust/syntaxes/rust.tmLanguage.json#L1018-L1031

[archseer]

You can use a scope inspector to see the scopes used, but I don't have VSCode installed: https://code.visualstudio.com/api/language-extensions/syntax-highlight-guide#textmate-tokens-and-scopes

[kirawi]

[archseer]

Hmm microsoft/vscode#126957

I also found this: https://code.visualstudio.com/api/language-extensions/semantic-highlight-guide#predefined-textmate-scope-mappings

Based on your screenshot: Looks like the "textmate scopes" is simply type.rust, I think the different styling you previously noticed for enum variants (outside of definitions) could be due to special casing Result/Option?

I mean in cases like MyEnum::Hello(1), outside of pub enum ... { ... }

[kirawi]

They're both styled as variable.other.enummember. I think you're referring to struct constructors? Though actually it seems like those are being highlighted as entity.name.type.struct.rust

[archseer]

Yeah that's what I meant, it looks like definitions have different scopes from constructors/uses in code. Similar to @function vs @variable.function.

Since we're already changing a couple scopes, let's also do:

• @character -> @constant.character
• @escape -> @constant.character.escape
• @number/@float to @constant.numeric.{integer, float, complex}.

That way we reduce the amount of top level scopes, making it easier to style in simple themes (you just need to define @constant, or @constant.numeric to style all numbers)

[kirawi]

Just need to update the book.

[archseer]

All the themes need to be updated to use these new scopes too.

[kirawi]

@raygervais I noticed that

"ui.popup" = { bg = "base01" }
"ui.window" = { bg = "base00" }
"ui.help" = { bg = "base01", fg = "base06" }

have duplicated keys in base16.

[kirawi]

How do I fix the submodule change? I don't know how that even happened.

[sudormrfbin]

How do I fix the submodule change? I don't know how that even happened.

Is it the change with elixir ? #875

[kirawi]

It's with cpp, presumably this happened when I did git pull upstream master

[archseer]

git sync; git submodule update --init?

[kirawi]

Doesn't seem like it. Maybe my case is different since I set upstream to this repo while origin points to my fork?

[archseer]

Well now that you committed and pushed the submodule update it considers that to be what it's syncing to maybe?

[archseer]

Ah sorry, was supposed to be git submodule sync too. Easiest fix would be to manually update the submodule to the same commit as in the repo

[kirawi]

How do you do that?

[kirawi]

Curse you git... curse you.

[archseer]

Maybe constant.numeric.integer? I know it's not necessarily correct, but if someone defines constant.numeric.integer and constant.numeric.float but not constant.numeric then this won't get highlighted.

[archseer]

Though I guess we should just recommend styling constant.numeric