Issue adding support for VHDL language

[m-kru]

I am trying to add support for the VHDL language. I have followed instruction from the https://docs.helix-editor.com/guides/adding_languages.html.

I have added following section to the languages.toml:

[[language]]
name = "vhdl"
scope = "source.vhdl"
injection-regex = "vhdl"
file-types = ["vhdl", "vhd"]
comment-token = "--"
indent = { tab-width = 3, unit = "   " }
roots = []

[[grammar]]
name = "vhdl"
source = { git = "https://github.com/alemuller/tree-sitter-vhdl", rev = "a3b2d84990527c7f8f4ae219c332c00c33d2d8e5" }

Then, I have recompiled and installed helix. I have also added highlights.scm to the ~/.config/helix/runtime/queries/vhdl.
However, when I open any .vhd file it looks like something is wrong, as there is no highlighting and tree-sitter-scopes command always returns [], no matter where in the source code the cursor is placed.

[the-mikedavis]

Are these the highlights you're using? https://github.com/alemuller/tree-sitter-vhdl/blob/main/queries/highlights.scm

It looks like they only highlight syntax errors. For general purpose syntax highlighting, it looks like you'll need to write the queries yourself; I don't see existing ones in nvim-treesitter to use as inspiration.

[m-kru]

Ok, it is all about writing highlights.scm now. Do you know any good tutorial on this? The tree-sitter documentation is kind of poor in this scope.

[the-mikedavis]

Actually I think the tree-sitter documentation is quite good. It splits up the query syntax and syntax highlighting docs into separate pages though and you should read both before writing queries.

I would recommend installing the tree-sitter CLI. Then in the tree-sitter-vhdl repository you can run tree-sitter generate to generate the C parser, tree-sitter build-wasm to generate the WASM bindings, and then tree-sitter playground to open an interactive web playground that shows you the full tree as you type. You can enable queries in that playground and try writing queries in that window against example VHDL documents for the fastest feedback loop.

[m-kru]

But where do I get a list of available classes/types of queries? For example

[
  "all"
  "begin"
] @keyword

What are other options than @keyword? Is it defined in the helix source code or is it defined by the tree-sitter?

[m-kru]

The syntax is quite simple, but I do not know what exactly can be defined, and I do not know where to look for.

[the-mikedavis]

For those scopes, we have a tree of them here: https://docs.helix-editor.com/master/themes.html#syntax-highlighting. Each consumer of tree-sitter tends to use sligthly different captures @keyword/@field/etc

Things in queries that are strings should be literal strings in the grammar file. For example, in the vdhl's _function_name rule, "function" is a literal string so you can capture than in a query with "function" @keyword. The square brackets allow you to do "alternation" - select competing patterns and have them captured by the same capture. So you could do

["(" ")"] @punctuation.bracket

(the parens are literals in the function_call rule)

And anything in the grammar that's an object key can be queried as a node (the syntax (node_name) for a key "node_name" - these are called "named nodes") except for "unnamed nodes" which is any key starting with an underscore. Those cannot be queried.

You can nest named nodes together in order to pattern match on that structure:

(function_call "(" (association_list) ")")

So say for example that association_list is valid under some other node than function_call. If you only want to capture association_list when it's under function_call then you would use the syntax

(function_call (association_list) @foo)

[m-kru]

@the-mikedavis do you know whether it is possible to query to highlight node when it matches some string?

I have a simple node:

    basic_identifier: $ => token(seq(
      /[a-zA-Z]/,
      repeat(seq(
        optional('_'),
        /[a-zA-Z0-9]/
      ))
    )),

This node should be highlighted only if it matches some string pattern, for example std_logic_vector. Is it doable without introducing an extra node and making a choice?

I have asked in the tree-sitter repo tree-sitter/tree-sitter#1838, but the answer is not unequivocal.

[the-mikedavis]

Yep, we implement the #match? predicate which can be used to match nodes via regular expression:

helix/runtime/queries/java/highlights.scm

Lines 27 to 32 in 8a75795

	((field_access
	object: (identifier) @type)
	(#match? @type "^[A-Z]"))
	((scoped_identifier
	scope: (identifier) @type)
	(#match? @type "^[A-Z]"))

In this case though since you know exactly what it should match to, you can also use the #eq? predicate:

((basic_identifier) @variable.builtin
 (#eq? @variable.builtin "std_logic_vector"))