Commit Sizing
Commit sizing categorizes commits by their total lines changed, providing insight into how developers structure their work.
Size Buckets
Total lines = lines added + lines removed.
| Size | Lines Changed | Typical Content |
|---|---|---|
| S (Small) | < 100 | Bug fixes, config changes, small features |
| M (Medium) | 100 – 500 | Standard features, meaningful refactors |
| L (Large) | 501 – 1,500 | Major features, significant refactors |
| XL (Extra Large) | > 1,500 | Migrations, initial implementations, generated code |
Complexity Smell Detection
In addition to size, commits are checked for a complexity smell:
A commit touching more than 8 files, regardless of line count, gets flagged as having a complexity smell. This catches wide-but-shallow changes that might indicate:
- Shotgun surgery (a change requiring edits across many files)
- Incomplete abstractions
- Cross-cutting concerns that should be centralized
Metrics
For each contributor over a date range:
| Metric | Description |
|---|---|
distribution | Count of commits per size bucket (S, M, L, XL) |
complexity_smells | Count of commits with > 8 files changed |
percentages | Percentage breakdown across buckets |
Healthy Patterns
Mostly S and M commits
A contributor whose work is primarily small and medium commits is likely:
- Practicing good commit discipline
- Breaking work into reviewable chunks
- Shipping incrementally
Frequent L and XL commits
Frequent large commits might indicate:
- Difficulty decomposing work (worth coaching on)
- Generated code or migrations (legitimate, context-dependent)
- Long-running branches merged all at once
- Infrequent committing habits
High complexity smell rate
Many commits touching 8+ files suggest:
- The codebase may need better abstractions
- Work is cross-cutting and tightly coupled
- Or the developer is making sweeping changes (refactors, renames)
Best practice
The ideal distribution is heavily weighted toward S and M, with occasional L commits for significant features and rare XL commits for migrations or initial setups. Use the distribution to guide conversations about work decomposition.