#!/usr/bin/env python3
"""CHECK: re-derive the stackchain open-issues table from the source snapshot
(issues-source.json) and diff against the committed artifact (issues.csv).

Exit 0 ONLY if the artifact is a faithful re-derivation. The comparison is
order-independent (both sides sorted by repo + number) so rendering order in the
CSV cannot mask or fake a mismatch.

This is a real regression test, not a tautology.
"""
import csv
import json
import sys

SRC_JSON = "/root/.hermes/wizard-outputs/ledger/issues-source.json"
ACTUAL_CSV = "/root/.hermes/wizard-outputs/ledger/issues.csv"

FIELDS = ["repo", "repo_id", "number", "title", "author", "state",
          "created_at", "updated_at", "labels"]


def canonical_key(r):
    return (r["repo"], int(r["number"]))


def reextract():
    src = json.load(open(SRC_JSON, encoding="utf-8"))
    rows = []
    for i in src.get("issues", []):
        row = {k: i.get(k) for k in FIELDS}
        # CSV stores identifiers as text; coerce here so the diff is type-clean.
        row["number"] = str(row["number"])
        row["repo_id"] = str(row["repo_id"])
        rows.append(row)
    return rows


def load_actual():
    with open(ACTUAL_CSV, encoding="utf-8", newline="") as f:
        return list(csv.DictReader(f))


def main():
    errors = []
    expected = reextract()
    actual = load_actual()

    if len(actual) != len(expected):
        errors.append(f"row count: actual={len(actual)} expected={len(expected)}")

    # Canonicalize both sides (order-independent) so rendering order cannot
    # disguise a content mismatch.
    exp_sorted = sorted(expected, key=canonical_key)
    act_sorted = sorted(actual, key=canonical_key)

    for i, (a, e) in enumerate(zip(act_sorted, exp_sorted), 1):
        for k in FIELDS:
            if a.get(k) != e[k]:
                errors.append(
                    f"row {i} ({e['repo']} #{e['number']}) field {k}: "
                    f"actual={a.get(k)!r} expected={e[k]!r}"
                )

    # Extra/missing safety: compare multisets via sorted keys explicitly.
    exp_keys = sorted((r["repo"], int(r["number"])) for r in expected)
    act_keys = sorted((r["repo"], int(r["number"])) for r in actual)
    if exp_keys != act_keys:
        missing = sorted(set(exp_keys) - set(act_keys))[:5]
        extra = sorted(set(act_keys) - set(exp_keys))[:5]
        errors.append(f"key sets differ; missing={missing} extra={extra}")

    if errors:
        print("CHECK FAILED:")
        for e in errors[:40]:
            print(" -", e)
        print(f"(showing up to 40 of {len(errors)} discrepancies)")
        sys.exit(1)

    print(f"CHECK OK: {len(actual)} open issues, re-derived from the source snapshot, "
          f"byte-identical to issues.csv")


if __name__ == "__main__":
    main()