#!/usr/bin/env bash
#
# show-review-delta.sh — Measure human review effort on AI-edited files.
#
# Compares the list of files the AI edited (from the impact log) against
# recent git commits to estimate how much the user modified AI output
# before committing.
#
# Usage: ./show-review-delta.sh [n_commits]
#   n_commits: number of recent commits to analyze (default: 10)

set -euo pipefail

PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$(cd "$(dirname "$0")/../.." && pwd)}"
LOG_FILE="$PROJECT_DIR/.claude/impact/impact-log.jsonl"
N_COMMITS="${1:-10}"

if [ ! -f "$LOG_FILE" ]; then
  echo "No impact log found at $LOG_FILE"
  exit 0
fi

cd "$PROJECT_DIR"

python3 -c "
import json, sys, subprocess, os

log_file = sys.argv[1]
n_commits = int(sys.argv[2])
project_dir = sys.argv[3]

# Collect all AI-edited files across all sessions (latest entry per session)
sessions = {}
with open(log_file) as f:
    for line in f:
        try:
            d = json.loads(line.strip())
            sid = d.get('session_id', '')
            if sid:
                sessions[sid] = d
        except Exception:
            pass

ai_files = {}  # file_path -> total edit count across sessions
for entry in sessions.values():
    for fp, count in entry.get('edited_files', {}).items():
        # Normalize to relative path
        rel = fp
        if rel.startswith(project_dir + '/'):
            rel = rel[len(project_dir) + 1:]
        ai_files[rel] = ai_files.get(rel, 0) + count

if not ai_files:
    print('No AI-edited files found in impact log.')
    print('(The edited_files field was added recently — older sessions lack it.)')
    sys.exit(0)

# Get recent commits and their changed files
result = subprocess.run(
    ['git', 'log', f'-{n_commits}', '--pretty=format:%H %s', '--name-only'],
    capture_output=True, text=True, cwd=project_dir
)

commits = []
current = None
for line in result.stdout.split('\n'):
    if not line.strip():
        continue
    if ' ' in line and len(line.split()[0]) == 40:
        # Commit line
        parts = line.split(' ', 1)
        current = {'hash': parts[0], 'msg': parts[1], 'files': [], 'ai_files': []}
        commits.append(current)
    elif current is not None:
        f = line.strip()
        current['files'].append(f)
        if f in ai_files:
            current['ai_files'].append(f)

print(f'=== Review Delta Analysis ({len(commits)} recent commits) ===')
print(f'AI-edited files tracked: {len(ai_files)}')
print()

total_files = 0
total_ai_overlap = 0

for c in commits:
    n_files = len(c['files'])
    n_ai = len(c['ai_files'])
    total_files += n_files
    total_ai_overlap += n_ai

    if n_ai > 0:
        marker = ' ← AI-touched'
    else:
        marker = ''
    print(f'  {c[\"hash\"][:8]} {c[\"msg\"][:60]}')
    print(f'    {n_files} files changed, {n_ai} were AI-edited{marker}')
    if n_ai > 0:
        for f in c['ai_files']:
            print(f'      {f} ({ai_files[f]} AI edits)')
    print()

if total_files > 0:
    overlap_pct = total_ai_overlap * 100 / total_files
else:
    overlap_pct = 0

print(f'=== Summary ===')
print(f'  Total files in commits:     {total_files}')
print(f'  Files also AI-edited:       {total_ai_overlap} ({overlap_pct:.0f}%)')
print(f'  Files only human-edited:    {total_files - total_ai_overlap}')
print()
if overlap_pct > 80:
    print('  High AI overlap — most committed code was AI-generated.')
    print('  Consider reviewing more carefully to maintain skill.')
elif overlap_pct > 40:
    print('  Moderate AI overlap — mixed human/AI contribution.')
elif total_ai_overlap > 0:
    print('  Low AI overlap — mostly human-written code committed.')
else:
    print('  No AI overlap detected in recent commits.')
" "$LOG_FILE" "$N_COMMITS" "$PROJECT_DIR"