Fallow - Codebase Health

Version updated for https://github.com/fallow-rs/fallow to version v2.38.0.

• This action is used across all versions by 7 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action, Fallow, is a high-performance, Rust-native codebase analyzer for TypeScript and JavaScript projects. It automates the detection of dead code, duplicate code, code complexity issues, and architectural problems without requiring any configuration, enabling developers to maintain cleaner, more efficient codebases. With sub-second analysis, it provides actionable insights, supports framework-specific plugins, and includes features like auto-removal previews and real-time file change monitoring.

What’s Changed

Highlights

Ownership risk analysis on hotspots: know who depends on which files, and surface single-contributor knowledge concentration before it becomes a retention problem. Runs entirely offline — no GitHub/GitLab API, no cloud — from your local git history.

fallow health --hotspots --ownership

● Hotspots (10 files, since 6 months)

  9/10 hotspots depend on a single recent contributor  ·  top authors: alice (6), bob (4)

   38.5 ▲  src/api/__mocks__/handlers.ts [test]
           6 commits    230 churn  0.19 density   1 fan-in  ▲ accelerating
         bus=1 (at risk)  top=alice (63%)

   20.9 ▲  src/wizards/ResultaatForm.tsx
           3 commits    357 churn  0.15 density   1 fan-in  ▲ accelerating
         bus=1 (sole author)  top=alice (100%)

   ...

Added

• fallow health --hotspots --ownership enriches each hotspot with:
  • bus_factor (Avelino truck factor — minimum contributors covering ≥50% of recency-weighted commits)
  • contributor_count after bot-pattern filtering
  • top_contributor with share, stale_days, commits, and a format discriminator
  • recent_contributors (up to three additional by share)
  • suggested_reviewers (first-class field: recent contributors with stale_days < 90 — paste-ready for AI agents doing PR review routing)
  • declared_owner from CODEOWNERS
  • Tristate unowned (true = no rule matches, false = rule matches, null = no CODEOWNERS file)
  • drift with human-readable drift_reason (fires when the original author has ≥30d file age AND <10% recent share)
• Human output leads with an ownership summary above the hotspot list so tech leads see the organizational pattern first (N/M hotspots depend on a single recent contributor · top authors: ...).
• Conditional severity: red + bold reserved for bus=1 (sole author) and bus=1 + accelerating cases so the marker stays meaningful on repos where most hotspots are single-contributor. The common bus=1 case is yellow; healthier bus=2+ is dimmed.
• Test-path tagging: hotspots in **/__tests__/**, **/__mocks__/**, *.test.*, *.spec.*, and **/test(s)/** get a [test] tag in human output and is_test_path: true in JSON. Test files stay in the ranking (test maintenance IS real work) but are visually distinguished.
• Ownership-aware JSON actions for AI agents:
  • low-bus-factor with file-specific candidate reviewers (Candidate reviewers: @ron.planken, @huub.sepers) when suggested_reviewers is non-empty
  • unowned-hotspot with a synthesized CODEOWNERS pattern (/src/api/users/) plus heuristic: "directory-deepest" so the strategy can evolve without breaking consumers
  • ownership-drift when the original author is no longer maintaining
• --ownership-emails={raw|handle|hash} controls author email privacy. Defaults to handle (local-part only, GitHub noreply unwrap: 12345+alice@users.noreply.github.com → alice). hash emits stable xxh3:<16hex> pseudonyms for regulated environments where author identities are sensitive in CI artifacts. raw for public OSS.
• health.ownership config section: botPatterns (glob patterns matched against raw author email; *noreply* deliberately omitted because it would filter the majority of real GitHub contributors) and emailMode.
• MCP check_health gains ownership: bool and ownership_email_mode: "raw"|"handle"|"hash" (typed enum, JSON Schema constrained).

Changed

• Churn layer now tracks per-author contributions. git log runs with --use-mailmap and the %at|%ae format to record authors. Per-file author contributions reference an interned email pool to keep the bitcode cache compact. Cache schema bumped to v2; older caches are automatically rejected on load.

Why this matters

On a 135 KLOC monorepo, running this against 6,662 commits of history surfaces:

• 9 of the top 10 hotspots depend on a single recent contributor
• Two dominant authors own 10 of 10 top hotspots
• No CODEOWNERS file means any of those people leaving is a retention crisis

The feature is deliberately raw-signals-first: fallow does not ship an opinionated “Knowledge Loss Risk” composite score. It gives you the numbers and lets you (or your AI agent) decide how to weight them. The actions[] array is the agent contract.

Research-backed: Avelino et al. on truck factor; Thongtanunam et al. (ICSE 2016) on ownership and code quality.

Community-requested via Discussion #116.

Full Changelog: https://github.com/fallow-rs/fallow/compare/v2.37.0...v2.38.0

• Fallow-Rs
• GitHub Actions