refactor: unified language registry — one source of truth, zero drift

[jamestexas]

Summary

Replaces 10 independent language/extension switch statements with a single internal/lang registry. Adding a new language now means editing ONE file.

• -45 net lines (580 added, 625 deleted) — added a whole package and still deleted more
• 9 bugs fixed across watcher, writeback, mount, config, and ingest dedup
• 18 languages fully supported everywhere (was 3-14 depending on which switch you hit)

What changed

File	Change
internal/lang/lang.go	NEW — Language struct, Registry, derived lookups (ForExt, ForName, ForPath, IsSourceExt, Extensions)
internal/lang/lang_test.go	NEW — exhaustive tests (completeness, aliases, case-insensitivity, no duplicate extensions)
internal/ingest/engine.go	langForExt + GetLanguage → thin wrappers over lang; processTreeSitterResult extracted (dedup)
internal/ingest/language.go	DetectLanguageFromExt → 4-line wrapper; deleted GetLanguageProfile, enrichHCLNode, LanguageProfile
internal/ingest/sitter_flatten.go	Uses lang.ForName().EnrichNode instead of deleted GetLanguageProfile
internal/ingest/watcher.go	Deleted sourceExtensions map; isSourceFile → lang.IsSourceExt()
internal/writeback/validate.go	LanguageForPath → 4-line wrapper over lang.ForPath() (was 8 langs, now 18)
cmd/schemas.go	presetSchemas derived from lang.Registry in init()
cmd/infer.go	sourceCodePresets derived from lang.Registry; uses lang.ForExt
cmd/mount.go	40-line infer switch → 3-line lang.ForExt() lookup
cmd/config.go	detectProjectType + sentinelFiles derived from lang.Registry (was 3 langs, now 18)
cmd/build.go	golang.GetLanguage() → lang.ForName("go").Grammar()
internal/linter/linter.go	golang.GetLanguage() → lang.ForName("go").Grammar()

Bugs fixed

1. Watcher missing 8 languages — Java/C/C++/Ruby/PHP/Kotlin/Swift/Scala file changes silently ignored
2. Writeback missing 10 languages — validation broken for most non-original languages
3. mount.go missing elixir, toml — --infer broken for these
4. detectProjectType missing 15 languages — only detected Go/Python/SQL
5. Broken file ID collision — sequential ingest used basename (collision), now SHA256
6. Missing _project_files fallback — sequential path silently dropped unmatched files
7. Missing RecordFile — sequential path broke incremental re-ingestion
8. HCL ref query under wrong key — registered as "hcl" but langForExt returns "terraform" (fixed in fix: watcher FD leak — respect .gitignore + unified skip list #129)
9. langForExt / DetectLanguageFromExt / GetLanguage — three copies of the same switch

Test plan

• internal/lang — 13 tests (completeness, all extensions, aliases, case-insensitivity, sorted output, no duplicates, EnrichNode)
• All existing tests pass unchanged
• grep -r 'case ".go"' internal/ cmd/ returns zero matches (no more switches)
• lang.ForName("hcl") returns terraform (backward compat alias)
• task check passes (fmt + vet + lint + test + validate)

Bead

mache-7gp

[Copilot]

Pull request overview

Refactors language detection/mapping across the codebase to use a single internal/lang registry, eliminating duplicated extension/name switch statements and aligning ingestion, watcher behavior, schema presets, and project detection on one canonical source of truth.

Changes:

• Introduces internal/lang with a Language registry and derived lookup helpers (ForExt, ForName, ForPath, IsSourceExt, Extensions) plus tests.
• Updates ingest pipeline, watcher, writeback validation, and AST flattening/enrichment to delegate language/grammar selection to internal/lang.
• Derives CLI preset schema lists and project-type detection sentinel files from the registry instead of hardcoded maps/switches.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
internal/lang/lang.go	New central language registry + lookup helpers; migrates HCL enrichment here.
internal/lang/lang_test.go	Adds tests to validate registry completeness, aliases, extensions, sorting, and enrichment hooks.
internal/ingest/engine.go	Replaces per-language switches with registry lookups; extracts processTreeSitterResult to dedupe logic.
internal/ingest/language.go	Collapses DetectLanguageFromExt into a thin wrapper over the registry; removes profile/enrichment types.
internal/ingest/sitter_flatten.go	Switches enrichment lookup to lang.ForName(...).EnrichNode.
internal/ingest/watcher.go	Replaces local sourceExtensions set with lang.IsSourceExt.
internal/writeback/validate.go	Delegates extension→grammar mapping to lang.ForPath(...).Grammar().
internal/linter/linter.go	Uses registry-provided Go grammar instead of direct tree-sitter import.
cmd/schemas.go	Builds presetSchemas from lang.Registry at init time.
cmd/infer.go	Builds source-code preset mapping from lang.Registry; uses registry in language detection and parsing.
cmd/mount.go	Replaces large infer switch with lang.ForExt lookup; updates call extractor to use registry.
cmd/config.go	Derives sentinel files and extension counting from lang.Registry for project-type detection.
cmd/build.go	Uses lang.ForName("go").Grammar() instead of direct Go grammar import.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.