"""Independent audit of the Stage B/C datasets (review point 7).

Deliberately separate from the generator code: this script knows nothing
about problem families or their planted bugs. It samples 50 examples,
re-executes each one's tests against the shipped answer/reference
(must PASS) and against a generically hand-broken version
(must FAIL — otherwise the tests are not discriminating).

Usage: python audit.py [pilot|full]
"""
import json
import os
import random
import sys

sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
from framework import run_tests, MASTER_SEED

WORK = os.path.dirname(os.path.abspath(__file__))


def hand_break(code):
    """Audit-side generic breaker — independent of family bug machinery."""
    candidates = []
    for a, b in [(" == ", " != "), (" < ", " > "), (" > ", " < "),
                 (" <= ", " >= "), (" >= ", " <= "), (" + ", " - "),
                 (" - ", " + "), (" * ", " // "), (" % ", " * ")]:
        if a in code:
            candidates.append(code.replace(a, b, 1))
    lines = code.splitlines()
    for i, ln in enumerate(lines):
        s = ln.lstrip()
        if s.startswith("return ") and s.strip() not in ("return None", "return"):
            m = lines[:]
            m[i] = ln[:len(ln) - len(s)] + "return None"
            candidates.append("\n".join(m))
            break
    for cand in candidates:
        if cand != code:
            try:
                compile(cand, "<audit>", "exec")
                return cand
            except SyntaxError:
                continue
    return None


def load(path):
    with open(path) as f:
        return [json.loads(ln) for ln in f if ln.strip()]


def main():
    mode = sys.argv[1] if len(sys.argv) > 1 else "full"
    pre = "pilot_" if mode == "pilot" else ""
    rows_b = load(os.path.join(WORK, f"{pre}stage_b.jsonl"))
    rows_c = load(os.path.join(WORK, f"{pre}stage_c.jsonl"))
    verify = {r["id"]: r for r in load(os.path.join(WORK, f"{pre}stage_c_verify.jsonl"))}

    rng = random.Random(MASTER_SEED)
    sample_b = rng.sample(rows_b, min(30, len(rows_b)))
    sample_c = rng.sample(rows_c, min(20, len(rows_c)))

    findings = []
    checked = passed_orig = failed_broken = 0
    for r in sample_b:
        checked += 1
        if run_tests(r["answer"], r["tests"]):
            passed_orig += 1
        else:
            findings.append({"id": r["id"], "set": "stage_b",
                             "issue": "shipped answer FAILS its own tests"})
        brk = hand_break(r["answer"])
        if brk is None:
            findings.append({"id": r["id"], "set": "stage_b",
                             "issue": "audit breaker produced no variant"})
        elif not run_tests(brk, r["tests"]):
            failed_broken += 1
        else:
            findings.append({"id": r["id"], "set": "stage_b",
                             "issue": "hand-broken answer still PASSES tests"})
    for r in sample_c:
        checked += 1
        ref = verify[r["id"]]["reference_solution"]
        if run_tests(ref, r["tests"]):
            passed_orig += 1
        else:
            findings.append({"id": r["id"], "set": "stage_c",
                             "issue": "reference FAILS its own tests"})
        brk = hand_break(ref)
        if brk is None:
            findings.append({"id": r["id"], "set": "stage_c",
                             "issue": "audit breaker produced no variant"})
        elif not run_tests(brk, r["tests"]):
            failed_broken += 1
        else:
            findings.append({"id": r["id"], "set": "stage_c",
                             "issue": "hand-broken reference still PASSES tests"})

    report = {
        "sampled": checked,
        "original_pass": passed_orig,
        "broken_fail": failed_broken,
        "findings": findings,
        "verdict": "PASS" if (passed_orig == checked and failed_broken == checked
                              and not findings) else "FINDINGS",
    }
    out = os.path.join(WORK, f"audit_report_{mode}.json")
    with open(out, "w") as f:
        json.dump(report, f, indent=2)
    print(f"[audit] sampled {checked}: originals pass {passed_orig}/{checked}, "
          f"broken fail {failed_broken}/{checked}, findings {len(findings)}")
    for fd in findings:
        print(f"[audit] FINDING {fd['set']} {fd['id']}: {fd['issue']}")
    print(f"[audit] verdict: {report['verdict']} -> {out}")

    # fold summary into the manifest
    mtag = "pilot" if mode == "pilot" else "v1"
    mp = os.path.join(WORK, f"manifest_{mtag}.json")
    if os.path.exists(mp):
        man = json.load(open(mp))
        man["audit"] = {"sampled": checked, "original_pass": passed_orig,
                        "broken_fail": failed_broken,
                        "findings": len(findings), "verdict": report["verdict"],
                        "report_file": os.path.basename(out)}
        json.dump(man, open(mp, "w"), indent=2)
        print(f"[audit] manifest updated: {mp}")


if __name__ == "__main__":
    main()