"""Recompute every headline number of F0003 from data/ and assert it matches the recorded values
(finding.yaml in the findings repo, expected_numbers.json in a downloaded copy).

    python check.py           # assert
    python check.py --print   # show computed values (for authoring)
"""

import csv
import json
import statistics as st
import sys
from pathlib import Path


HERE = Path(__file__).parent


def pct(k, n):
    return round(100 * k / n, 1)


def compute():
    rows = list(csv.DictReader(open(HERE / "data" / "scores.csv", newline="")))
    out = {}
    for s, tag in (("reaction_qm", "rqm"), ("geodesic", "geo")):
        r = [x for x in rows if x["set"] == s]
        by = {o: [x for x in r if x["outcome"] == o] for o in ("lower", "matched", "higher", "no_channel")}
        withch = len(r) - len(by["no_channel"])
        lo = [float(x["lower_by_kcal"]) for x in by["lower"]]
        out.update({
            f"{tag}_n": len(r),
            f"{tag}_with_channel": withch,
            f"{tag}_lower": len(by["lower"]),
            f"{tag}_matched": len(by["matched"]),
            f"{tag}_higher": len(by["higher"]),
            f"{tag}_no_channel": len(by["no_channel"]),
            f"{tag}_lower_pct_of_set": pct(len(by["lower"]), len(r)),
            f"{tag}_lower_pct_of_with_channel": pct(len(by["lower"]), withch),
            f"{tag}_lower_single_step": sum(x["best_kind"] == "channel" for x in by["lower"]),
            f"{tag}_lower_multistep": sum(x["best_kind"] == "alternate" for x in by["lower"]),
            f"{tag}_lower_median_kcal": round(st.median(lo), 2),
            f"{tag}_lower_max_kcal": round(max(lo), 2),
            f"{tag}_not_finished": sum(x["returncode"] != "0" for x in r),
        })
    rqm = [x for x in rows if x["set"] == "reaction_qm"]
    for b in ("1", "2"):
        rb = [x for x in rqm if x["batch"] == b]
        out[f"rqm_batch{b}_n"] = len(rb)
        out[f"rqm_batch{b}_lower"] = sum(x["outcome"] == "lower" for x in rb)
    return out


def recorded_numbers() -> dict:
    if (HERE / "finding.yaml").exists():
        import yaml
        return yaml.safe_load((HERE / "finding.yaml").read_text())["evidence"]["numbers"]
    return json.loads((HERE / "expected_numbers.json").read_text())


if __name__ == "__main__":
    computed = compute()
    if "--print" in sys.argv:
        print(json.dumps(computed, indent=1))
        sys.exit()
    numbers = recorded_numbers()
    bad = [f"{k}: computed {computed.get(k)!r}, recorded {v!r}" for k, v in numbers.items() if computed.get(k) != v]
    if bad:
        sys.exit("mismatch:\n  " + "\n  ".join(bad))
    print(f"ok: {len(numbers)} numbers recomputed from data/")