Commit Parser

Commit Parser — Commit Message Parsing and Structuring Methodology

A skill that enhances the commit analysis capabilities of the commit-analyst.

Target Agents

• commit-analyst — Extracts meaningful change information from git history
• change-classifier — Classifies changes using parsed results

Conventional Commits Parsing Rules

Regex Pattern

^(?<type>feat|fix|docs|style|refactor|perf|test|ci|chore|build|revert)
(?:\((?<scope>[^)]+)\))?
(?<breaking>!)?
:\s*
(?<subject>.+)$

Body: BREAKING CHANGE: <description>
Footer: Refs: #123, Closes #456

Parsed Result Structure

ParsedCommit = {
    "hash": "abc1234",
    "type": "feat",           # Change type
    "scope": "auth",          # Impact scope (optional)
    "breaking": True/False,   # Breaking Change
    "subject": "add OAuth2",  # Subject
    "body": "...",            # Body
    "footer": {
        "refs": ["#123"],     # Referenced issues
        "closes": ["#456"],   # Closed issues
        "breaking_change": "" # BREAKING CHANGE description
    },
    "author": "name <email>",
    "date": "2025-01-15",
    "files_changed": ["src/auth.py"],
    "insertions": 50,
    "deletions": 10
}

Non-Conventional Commit Classification Strategy

For projects that do not use Conventional Commits:

Keyword-Based Classification

classification_rules = {
    "feat": ["add", "new", "implement", "introduce", "support"],
    "fix": ["fix", "bug", "patch", "resolve", "correct", "hotfix"],
    "refactor": ["refactor", "restructure", "reorganize", "clean"],
    "perf": ["performance", "optimize", "speed", "faster", "cache"],
    "docs": ["doc", "readme", "comment"],
    "test": ["test", "spec", "coverage"],
    "ci": ["ci", "pipeline", "workflow", "deploy"],
    "style": ["format", "lint", "whitespace"],
    "chore": ["bump", "update dep", "upgrade", "dependency"]
}

Diff-Based Classification (when keyword matching fails)

1. Analyze changed file extensions:
   - .md/.rst -> docs
   - test_*.py / *.test.js -> test
   - .yml/.yaml (CI paths) -> ci

2. Analyze change patterns:
   - New file added + export -> feat
   - Existing file modified + error/exception -> fix
   - Code moved without new imports -> refactor

3. Analyze change scale:
   - +500 lines or more -> likely feat
   - +/-10 lines or fewer -> likely fix/style

PR/Issue Mapping Strategy

# Extract issue numbers from commits
issue_patterns = [
    r'#(\d+)',                    # #123
    r'(?:close|fix|resolve)s?\s+#(\d+)',  # closes #123
    r'([A-Z]+-\d+)',              # JIRA-123
    r'GH-(\d+)',                  # GH-123
]

# Extract related issues from PRs (Merge commit)
merge_pattern = r'Merge pull request #(\d+)'

# Extract PR number from squash merges
squash_pattern = r'\(#(\d+)\)$'

Change Impact Score Calculation

def impact_score(commit):
    score = 0

    # Based on file count
    score += min(len(commit.files), 10) * 2  # Max 20

    # Based on lines changed
    changes = commit.insertions + commit.deletions
    if changes > 500: score += 30
    elif changes > 100: score += 20
    elif changes > 30: score += 10
    else: score += 5

    # Breaking Change
    if commit.breaking: score += 50

    # Public API changes
    if any('api' in f or 'public' in f for f in commit.files):
        score += 15

    # Database migrations
    if any('migration' in f for f in commit.files):
        score += 20

    return min(score, 100)

# Score interpretation:
# 80-100: Critical — Release note highlight
# 50-79: High — Include in release notes
# 20-49: Medium — Include in change list
# 0-19: Low — Include in detailed list only

Git Command Recipes

# Commits between two tags
git log v1.2.0..v2.0.0 --oneline --no-merges

# Include changed files
git log v1.2.0..v2.0.0 --name-only --pretty=format:"%H|%s|%an|%ai"

# Diff statistics
git diff v1.2.0..v2.0.0 --stat

# Contributor list
git log v1.2.0..v2.0.0 --format="%aN" | sort -u

# Breaking Change commits only
git log v1.2.0..v2.0.0 --grep="BREAKING" --grep="!" --all-match