fix: references/rename for struct_init_one, struct_init_one_comma

[WillLillis]

This fixes some cases where references to fields in some struct initializations aren't picked up, causing features like renaming and reference viewing to fail. Currently the fix only works one way, in that a rename invoked on the actual struct initialization will still fail, but will work when invoked on other locations such as the struct's definition or a field access. I took some time to try and fix the other direction, but I don't think I'm familiar enough with the codebase to do that yet. I added a test case for this but left it commented out as it currently fails.

This fix only covers the cases of .struct_init_one and .struct_init_one_comma. I'd really like to look into fixing the other node cases (e.g. .struct_init, .struct_init_comma, etc. if I'm understanding things properly) but wanted to cover this first case and get some feedback/ a sanity check before proceeding on. Here's a quick demo showing both renaming and references with the improvements:

Related Issues: #1837, #1700

[Techatrix]

It should be possible to handle all the other .struct_init_* cases by doing something like this:

zls/src/features/inlay_hints.zig

Lines 551 to 556 in 93b7bbd

	var buffer: [2]Ast.Node.Index = undefined;
	const struct_init = tree.fullStructInit(&buffer, node).?;
	for (struct_init.ast.fields) |value_node| { // the node of `value` in `.name = value`
	const name_token = tree.firstToken(value_node) - 2; // math our way two token indexes back to get the `name`
	const name_loc = offsets.tokenToLoc(tree, name_token);
	const name = offsets.locToSlice(tree.source, name_loc);

The following patch appears to fix the "other direction" that previously didn't work:

diff --git a/src/features/references.zig b/src/features/references.zig
index d4b665d1..09646b4d 100644
--- a/src/features/references.zig
+++ b/src/features/references.zig
@@ -443,6 +443,7 @@ pub fn referencesHandler(server: *Server, arena: std.mem.Allocator, request: Gen
             break :z null;
         },
         .label => try Analyser.getLabelGlobal(source_index, handle, name),
+        .enum_literal => try analyser.getSymbolEnumLiteral(arena, handle, source_index, name),
         else => null,
     } orelse return null;

[WillLillis]

It should be possible to handle all the other .struct_init_* cases by doing something like this:

Awesome, I'll take a pass at the other cases soon. Thank you! :)

Edit: Should those additions for the other cases be pushed on top of this PR, or put into a new one? That was straightforward enough I'll just assume for now it's ok to push on top of the existing PR.

[WillLillis]

Thought I ran zig fmt before pushing last night, really sorry about that!

[WillLillis]

Unless I'm mistaken, it looks like CI is failing now due to the recent deprecations in ziglang/zig@f97c2f2, which is fixed in #1904?

[Techatrix]

I removed the initial switch case that just handled .struct_init_one.* because the other case is already capable of handling all variants. Otherwise, this looks great!

[WillLillis]

I removed the initial switch case that just handled .struct_init_one.* because the other case is already capable of handling all variants. Otherwise, this looks great!

Makes sense, thanks so much for the help!