analyze-code-churn

Analyze Churn

Analyze code churn and thrash patterns in a git repository. Produces a structured report with hotspot detection, rework vs. refactor classification, temporal coupling analysis, and severity-rated findings with investigation recommendations.

Input

$ARGUMENTS = analysis scope — time window (e.g., "30 days", "last sprint", "since v2.1"), optional path filter (e.g., "src/api/"), and optional focus area (e.g., "why is auth churning", "sprint-over-sprint comparison").

Process

1. Parse Scope

Extract from $ARGUMENTS:

• Time window: Default to 30 days if not specified. Accept natural language ("last sprint", "since January") and convert to --since/--after git flags.
• Path filter: If a directory or file pattern is specified, scope all git commands to that path. Default to the entire repository.
• Focus area: If a specific question is asked, tailor the analysis toward answering it. Otherwise, produce the full report.

If $ARGUMENTS is empty, use a 30-day window across the entire repository.

2. Establish Baseline

Run the baseline period (the window immediately preceding the analysis window) to enable trend comparison. For a 30-day analysis window, the baseline is the prior 30 days.

Gross churn (analysis window):

git log --numstat --format="" --no-merges --since="<start>" -- <path> \
  | awk '{a+=$1; d+=$2} END {print "added:", a, "deleted:", d, "gross:", a+d}'

Gross churn (baseline window):

git log --numstat --format="" --no-merges --since="<baseline_start>" --until="<start>" -- <path> \
  | awk '{a+=$1; d+=$2} END {print "added:", a, "deleted:", d, "gross:", a+d}'

Commit counts (both windows):

git rev-list --count --no-merges --since="<start>" HEAD -- <path>
git rev-list --count --no-merges --since="<baseline_start>" --until="<start>" HEAD -- <path>

Record the baseline metrics for trend comparison in Step 7.

3. Detect Hotspots

Identify files with the highest change frequency, weighted by size as a complexity proxy.