"""Hit table for Meilisearch: keyword, vector-only (semanticRatio 1), hybrid at 0.5 (the documented default), plus a ratio sweep."""
import json
from common import EXACT, FETCH, TOP, hit_rank, meili, pages, qvec, questions

QS = questions() + [EXACT]


def search(q, ratio=None, vec=True, filt=None, key=None, extra=None):
    body = {"q": q, "limit": FETCH, "attributesToRetrieve": ["url", "heading"]}
    if ratio is not None:
        body["hybrid"] = {"embedder": "nomic", "semanticRatio": ratio}
        body["vector"] = qvec(q)
    if filt:
        body["filter"] = filt
    body.update(extra or {})
    return meili("POST", "/indexes/docs/search", body, key=key)


def urls(res):
    return [h["url"] for h in res["hits"]]


def vector_only(q):  # q is passed too; at ratio 1.0 only the vector matters
    return search(q, 1.0)


if __name__ == "__main__":
    print("keyword settings: defaults (matchingStrategy last, typoTolerance on, prefix search on); fetch", FETCH, "chunks, top", TOP, "pages")
    print("modes: keyword = no hybrid; vector = semanticRatio 1.0; hybrid = semanticRatio 0.5 (docs default)")
    modes = {"keyword": lambda q: search(q), "vector": lambda q: search(q, 1.0), "hybrid": lambda q: search(q, 0.5)}
    print(f"{'question':<70} " + " ".join(f"{m:>8}" for m in modes))
    tot = {m: 0 for m in modes}
    tot6 = {m: 0 for m in modes}
    top5 = {}
    for i, q in enumerate(QS):
        row = []
        for m, fn in modes.items():
            res = fn(q["question"])
            r = hit_rank(urls(res), q["expected_url_contains"])
            tot[m] += r is not None
            tot6[m] += r is not None and i < 6
            row.append(f"{r or '-':>8}")
            if m == "vector":
                top5[q["question"]] = pages(urls(res))[:TOP]
        print(f"{q['question'][:68]:<70} " + " ".join(row))
    print("hits x/6 (six questions):", {m: f"{v}/6" for m, v in tot6.items()})
    print("429 rank:", {m: hit_rank(urls(fn('429')), EXACT['expected_url_contains']) for m, fn in modes.items()})
    json.dump(top5, open("meili_vector_top5.json", "w"), indent=1)

    print("\n--- semanticRatio sweep: rank of expected page (- = miss), last column = x/6 hits, then 429 rank ---")
    for ratio in (0.0, 0.25, 0.5, 0.75, 1.0):
        ranks = [hit_rank(urls(search(q["question"], ratio)), q["expected_url_contains"]) for q in QS]
        print(f"ratio {ratio:<5}", " ".join(f"{r or '-':>2}" for r in ranks[:6]), f"| {sum(r is not None for r in ranks[:6])}/6 | 429: {ranks[6] or '-'}")
    print("\n--- hybrid semanticHitCount at ratio 0.5 (of", FETCH, "requested) ---")
    for q in QS:
        res = search(q["question"], 0.5)
        print(f"{q['question'][:60]:<62} semanticHitCount {res.get('semanticHitCount')} estimatedTotalHits {res.get('estimatedTotalHits')}")