mirror of
https://github.com/colbymchenry/codegraph.git
synced 2026-10-19 06:22:38 +08:00
Colby McHenry
59b8de21fd
New standalone RazorExtractor (follows the svelte/vue/liquid pattern) links
markup-driven C# that previously had no static in-repo caller:
- `@model Foo` / `@inherits Bar<Foo>` → the view-model / base type (.cshtml + .razor)
- `@inject IService svc` → the injected service type
- `@typeof(X)` → the referenced type
- Blazor `<MyComponent/>` tags + generic `TItem="Foo"` args → the component class
Wiring: `.cshtml`/`.razor` → new `razor` language (types.ts, grammars.ts
EXTENSION_MAP + isLanguageSupported/isGrammarLoaded + GrammarLanguage exclude +
display name); dispatch in tree-sitter.ts `extractFromSource`.
Key risks (per docs/design/template-markup-parser.md) all handled:
1. FAMILY GATE — `razor`+`csharp` share a new `dotnet` family in name-matcher,
so `@model Foo`→`Foo.cs` isn't dropped by the cross-family gate (082353e).
2. PASCALCASE vs HTML — only `[A-Z]`-initial tags are components (HTML is
lowercase); known Blazor framework components are skipped.
3. NODE DISCIPLINE — exactly one `component` node per file; tags → `references`
edges, never nodes (no per-tag explosion).
4. INDEXING — `.cshtml`/`.razor` registered so they're indexed.
Validated: eShopOnWeb — 61 razor nodes, 97 resolved markup→code edges
(`App.razor`→MainLayout, `Create.razor`→CatalogBrand/ICatalogItemService,
ToastComponent covered). FAIR coverage 77.2% → 79.9% (.cshtml/.razor counted as
view entries in the metric). No regression (gin/requests/Express unchanged,
cs-mediatr/cs-polly fine). Full suite 1175; 1 regression test.
Modest gain on this app is honest: residual DTOs are a name-collision
(`CatalogBrand` exists as both entity and DTO — edge resolves to the entity),
and many DTOs/services are used inside Blazor `@code { }` blocks not yet
delegated to the C# extractor — both documented as next increments in the scope.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>