Add Bash script support

[Omar-Elwazeery]

Summary

Fixes #1203.

• Register Bash/shell as a supported Tree-sitter language using the existing WASM grammar pipeline.
• Add Bash detection for .sh, .bash, .bats, and extensionless scripts with Bash/POSIX shell shebangs.
• Add a Bash extractor for shell functions, assignments, exported/readonly variables, command calls, and static source/script imports.
• Include extensionless Bash shebang scripts in scan/sync flows and add Bash import resolution for shell/Bats files.
• Add regression coverage for detection, extraction, scanning, sync, and .bats import resolution.

Testing

• ✅ npm test -- __tests__/extraction.test.ts __tests__/resolution.test.ts __tests__/sync.test.ts -t Bash
• ✅ npm run build
• ⚠️ npm test currently times out in the full-suite parallel run at __tests__/mcp-daemon.test.ts > proxy survives the daemon dying mid-session and keeps serving (#662). I confirmed the same full-suite timeout occurs on a clean base after stashing this branch, and the exact test passes when run isolated both before and after this change.

[Adam0120]

I tested this PR against a deployment repository with 101 Bash scripts and found a reproducible grammar ABI failure on common case ... esac statements (resolved is not a function), plus two accuracy issues around shell path semantics and canonical script-directory variables.

I opened a focused patch against the author's branch: Omar-Elwazeery#1

The patch vendors the official ABI-15 tree-sitter-bash 0.25.1 WASM with its MIT notice, keeps Bash imports exact-path, conservatively folds only a root-level deterministic script-directory idiom, and adds regressions. On a main-based integration branch it indexed all 101/101 .sh files and recovered the real cross-file source/call chains. Build and Bash-focused tests pass on the PR branch; full extraction + sync suites pass 615/615 on the main-based integration branch.

[Omar-Elwazeery]

Rebased onto the latest main and force-pushed (c8a48a2). Bash support is still verified with targeted Bash tests and a successful build. The PR is mergeable; the only remaining blocker is review approval.

[colbymchenry]

CodeGraph review

@Omar-Elwazeery — 1 finding on this change.

Overall risk: 🟠 Medium

This change adds Bash language support, including shebang detection, grammar loading, and a new extractor. The most important risk is that extensionless files are unconditionally classified as Bash for grammar loading, which could cause parsing failures for non‑Bash files.

Findings

Severity	Finding	Where
🟠 Medium	Overbroad language detection for extensionless files	src/extraction/index.ts:150 — see inline comment

Worth double-checking

• detectLanguageForGrammarLoad — Verify that treating any extensionless file as Bash (line 150) does not cause incorrect grammar loading for non‑Bash files.
• collectIncludedFiles — Note that collectIncludedFiles has no test coverage; consider adding tests to ensure it correctly filters files, especially with the new Bash shebang logic.

Business rules (2 honored · 2 not applicable)

No rules violated or at risk.

QA checklist — 3 things to verify in the running product

• cli — Run codegraph sync on a repository containing a Bash script with a proper shebang and confirm the script is indexed and appears in the language list. (Ensures the new Bash extraction and shebang detection work end‑to‑end.)
• cli — Run codegraph sync on a repository without any Bash scripts and verify the sync completes without errors. (Confirms that the added Bash support does not break existing sync behavior for other languages.)
• api — Query getSupportedLanguages() after sync and check that 'bash' is listed. (Validates that the language registry correctly includes Bash.)

Blast radius: 53 files affected beyond the diff · 161 symbols · 42 test files selected

Tests to run:

• __tests__/adaptive-explore-sizing.test.ts
• __tests__/android-res-exclusion.test.ts
• __tests__/concurrent-locking.test.ts
• __tests__/config-secret-redaction.test.ts
• __tests__/context-ranking.test.ts
• __tests__/context.test.ts
• __tests__/db-reopen-on-replace.test.ts
• __tests__/dynamic-boundaries.test.ts
• __tests__/exclude-config.test.ts
• __tests__/explore-blast-radius.test.ts
• __tests__/explore-corroboration-ranking.test.ts
• __tests__/explore-nl-stopword-collision.test.ts
• __tests__/explore-output-budget.test.ts
• __tests__/explore-result-count.test.ts
• __tests__/explore-synth-constant-endpoints.test.ts
• __tests__/extension-mapping.test.ts
• __tests__/extraction.test.ts
• __tests__/frameworks.test.ts
• __tests__/frontload-hook.test.ts
• __tests__/grammar-wasm-bytes.test.ts

---

[colbymchenry]

🟠 Medium — Overbroad language detection for extensionless files

detectLanguageForGrammarLoad treats any extensionless file as Bash, potentially loading the Bash grammar for non‑Bash files and causing parse errors (src/extraction/index.ts line 150).