tools/analyze_failures.py

#!/usr/bin/env python3
"""Analyze failure patterns in COMPREHENSIVE_SEMANTIC_AUDIT.md"""

import re
from collections import defaultdict


def analyze_failures(report_path: str):
    """Parse the audit report and categorize failures."""

    with open(report_path, "r", encoding="utf-8") as f:
        content = f.read()

    # Parse failure patterns
    failures = []
    for line in content.split("\n"):
        if "❌ FAIL" in line:
            # Extract card, ability, and error details
            match = re.search(r"\| ([^|]+) \| (Ab\d+) \| ❌ FAIL \| (.+) \|", line)
            if match:
                card = match.group(1).strip()
                ability = match.group(2).strip()
                details = match.group(3).strip()
                failures.append({"card": card, "ability": ability, "details": details})

    # Categorize failures
    categories = defaultdict(list)

    for fail in failures:
        details = fail["details"]

        # Extract the effect type and mismatch
        if "COST: DISCARD_HAND" in details:
            categories["COST_DISCARD_HAND"].append(fail)
        elif "COST: PAY_ENERGY" in details:
            categories["COST_PAY_ENERGY"].append(fail)
        elif "COST: TAP_MEMBER" in details:
            categories["COST_TAP_MEMBER"].append(fail)
        elif "EFFECT: BOOST_SCORE" in details:
            categories["EFFECT_BOOST_SCORE"].append(fail)
        elif "EFFECT: DRAW" in details:
            categories["EFFECT_DRAW"].append(fail)
        elif "EFFECT: ADD_BLADES" in details:
            categories["EFFECT_ADD_BLADES"].append(fail)
        elif "EFFECT: ADD_HEARTS" in details:
            categories["EFFECT_ADD_HEARTS"].append(fail)
        elif "EFFECT: TAP_OPPONENT" in details:
            categories["EFFECT_TAP_OPPONENT"].append(fail)
        elif "EFFECT: RECOVER_MEMBER" in details:
            categories["EFFECT_RECOVER_MEMBER"].append(fail)
        elif "EFFECT: RECOVER_LIVE" in details:
            categories["EFFECT_RECOVER_LIVE"].append(fail)
        elif "EFFECT: LOOK_AND_CHOOSE" in details:
            categories["EFFECT_LOOK_AND_CHOOSE"].append(fail)
        elif "EFFECT: MOVE_TO_DISCARD" in details:
            categories["EFFECT_MOVE_TO_DISCARD"].append(fail)
        elif "EFFECT: MOVE_TO_DECK" in details:
            categories["EFFECT_MOVE_TO_DECK"].append(fail)
        elif "DECK_SEARCH" in details:
            categories["DECK_SEARCH"].append(fail)
        else:
            categories["OTHER"].append(fail)

    # Print summary
    print("=" * 60)
    print("FAILURE ANALYSIS SUMMARY")
    print("=" * 60)
    print(f"\nTotal Failures: {len(failures)}")
    print("\nBy Category:")
    print("-" * 40)

    for cat, items in sorted(categories.items(), key=lambda x: -len(x[1])):
        print(f"{cat}: {len(items)} failures")

    print("\n" + "=" * 60)
    print("DETAILED BREAKDOWN")
    print("=" * 60)

    for cat, items in sorted(categories.items(), key=lambda x: -len(x[1])):
        print(f"\n### {cat} ({len(items)} failures)")
        print("-" * 40)

        # Group by error pattern
        error_patterns = defaultdict(list)
        for item in items:
            # Extract expected vs got
            match = re.search(r"Exp (-?\d+), Got (-?\d+)", item["details"])
            if match:
                exp, got = match.group(1), match.group(2)
                pattern = f"Exp {exp}, Got {got}"
                error_patterns[pattern].append(item)
            else:
                error_patterns["Other"].append(item)

        for pattern, pattern_items in sorted(error_patterns.items(), key=lambda x: -len(x[1])):
            print(f"  {pattern}: {len(pattern_items)} cases")
            # Show first 3 examples
            for ex in pattern_items[:3]:
                print(f"    - {ex['card']} {ex['ability']}")
            if len(pattern_items) > 3:
                print(f"    ... and {len(pattern_items) - 3} more")


if __name__ == "__main__":
    analyze_failures("reports/COMPREHENSIVE_SEMANTIC_AUDIT.md")