Allow limiting semantic token highlighting to identifiers

[Vexu]

Semantic token highlighting is nice and all but using a syntax file like the one provided by vscode-zig gives much more control. Limiting semantic token highlighting to only identifiers gives the best of both worlds. I think the C/C++ language server does something similar since the result looks much more like my what I was able to achieve with a quick patch.

Zls master version

With patch applied

Quick patch to enable this behavior, not sure how correct this is

diff --git a/src/semantic_tokens.zig b/src/semantic_tokens.zig
index eae2a54..8c813f3 100644
--- a/src/semantic_tokens.zig
+++ b/src/semantic_tokens.zig
@@ -66,6 +66,18 @@ const Builder = struct {
     }
 
     fn add(self: *Builder, token: ast.TokenIndex, token_type: TokenType, token_modifiers: TokenModifiers) !void {
+        switch (token_type) {
+            .type,
+            .parameter,
+            .variable,
+            .enumMember,
+            .field,
+            .errorTag,
+            .function,
+            .label, => {},
+            else => return,
+        }
+
         const tree = self.handle.tree;
         const starts = tree.tokens.items(.start);
         const tags = tree.tokens.items(.tag);
@@ -159,6 +171,18 @@ const Builder = struct {
     }
 
     fn addDirect(self: *Builder, tok_type: TokenType, tok_mod: TokenModifiers, start: usize, length: usize) !void {
+        switch (tok_type) {
+            .type,
+            .parameter,
+            .variable,
+            .enumMember,
+            .field,
+            .errorTag,
+            .function,
+            .label, => {},
+            else => return,
+        }
+
         const delta = offsets.tokenRelativeLocation(
             self.handle.tree,
             self.previous_position,

[SuperAuguste]

Quite the interesting idea, perhaps worthy of a configuration option :)

[Techatrix]

implemented by #1064