CSharp Language Support Plan

Implementation Plan: C# Code Indexing

Branch: 110-csharp-code-indexing | Date: 2026-01-27 | Spec: spec.md Input: Feature specification from .speckit/features/110-csharp-code-indexing/spec.md

Summary

Add C# as the 11th supported language in doc-serve's code indexing pipeline. This is a slot-in addition following the established pattern from 10 existing languages: add file extension mappings, tree-sitter parser setup, AST query patterns for symbol extraction, and content-based detection patterns. No new dependencies, no API changes, no architectural modifications.

Technical Context

Language/Version: Python 3.10+ Primary Dependencies: tree-sitter-language-pack (existing, includes c_sharp grammar), LlamaIndex CodeSplitter (existing) Storage: ChromaDB (existing vector store), disk-based BM25 index (existing) Testing: pytest (with async support) Target Platform: macOS/Linux local development Project Type: Monorepo (doc-serve-server, doc-svr-ctl, doc-serve-skill) Performance Goals: C# indexing performance on par with existing languages (~100 files/sec) Constraints: No new external dependencies; no API changes Scale/Scope: Standard C# projects (1-10k files)

Constitution Check

GATE: Must pass before Phase 0 research. Re-check after Phase 1 design.

I. Monorepo Modularity — PASS

• All changes are within doc-serve-server (the server package)
• No changes to doc-svr-ctl or doc-serve-skill
• No cross-package dependency changes

II. OpenAPI-First — PASS (No API Changes)

• FR-009 explicitly requires no API changes
• Existing /index and /query endpoints work unchanged
• The language=csharp filter uses the existing language query parameter

III. Test-Alongside — PASS

• Unit tests for C# language detection, AST parsing, and symbol extraction
• Integration tests for end-to-end C# indexing and querying
• Tests follow existing patterns from other language test files

IV. Observability — PASS

• C# chunks counted in existing code_chunks metric (FR-008)
• Existing structured logging covers C# files through the standard pipeline
• No new observability requirements

V. Simplicity — PASS

• ~100 lines of new code (well under 500-line threshold)
• 0 new dependencies
• 3 files modified (document_loader.py, chunking.py, and tests)
• Follows exact pattern of existing languages — no new abstractions

Project Structure

Documentation (this feature)

.speckit/features/110-csharp-code-indexing/
├── plan.md              # This file
├── research.md          # Phase 0 output (6 decisions)
├── quickstart.md        # Phase 1 output
└── tasks.md             # Phase 2 output (/speckit.tasks)

Source Code (repository root)

doc-serve-server/
├── doc_serve_server/
│   ├── indexing/
│   │   ├── document_loader.py   # MODIFIED: add .cs/.csx extensions, csharp content patterns
│   │   └── chunking.py          # MODIFIED: add c_sharp parser setup, C# AST query patterns
│   └── ...
└── tests/
    └── unit/
        ├── test_document_loader.py  # MODIFIED: add C# extension and content detection tests
        └── test_chunking.py         # MODIFIED: add C# AST parsing and symbol extraction tests

Structure Decision: Monorepo structure preserved. Only 2 source files modified, plus their existing test files. No new modules or packages created.

Implementation Phases

Phase 1: C# Language Support (US1 + US2 + US3 — All Stories)

Goal: Full C# indexing with AST-aware chunking, symbol metadata extraction, and content detection.

This feature is small enough to implement in a single phase. All three user stories touch the same two files and are tightly coupled.

1. Add .cs and .csx to EXTENSION_TO_LANGUAGE dict in document_loader.py
2. Add .cs and .csx to CODE_EXTENSIONS set in document_loader.py
3. Add csharp content-detection patterns to CONTENT_PATTERNS dict in document_loader.py
4. Add "c_sharp": "c_sharp" mapping to _setup_language() in chunking.py
5. Add C# AST query patterns to _get_symbols() in chunking.py — query for class_declaration, method_declaration, constructor_declaration, interface_declaration, property_declaration, enum_declaration, struct_declaration, record_declaration
6. Add XML doc comment extraction logic for C# /// comments in chunking.py
7. Write unit tests for all new code
8. Run task before-push to verify all quality gates

Files Modified: document_loader.py, chunking.py, test_document_loader.py, test_chunking.py

Key Design Decisions

Decision	Choice	Rationale
Grammar source	tree-sitter-language-pack c_sharp	Already a dependency; no new package needed
File extensions	.cs, .csx	Standard C# source and script extensions
Tree-sitter ID	c_sharp	Language pack naming convention
Symbol types	9 node types (class, method, constructor, interface, property, enum, struct, record, namespace)	Covers all common C# declarations
Doc comments	Extract /// comment text	Equivalent to Python docstrings and Java Javadoc
Content patterns	5 regex patterns (using System, namespace, property accessors, attributes, type declarations)	Disambiguates C# from Java

See research.md for full decision records with alternatives considered.