"""Verify every intended reference against Crossref (public API, polite pool) and build the verified bibliography.

For each seed (author words, title, year) we query Crossref, keep the best candidate only if the title similarity is
high and the year matches (±1), and record DOI, container title, volume, pages. Unverified entries are listed in
literature/unverified.csv and are NOT cited in the paper.
"""
import time, re, json, requests, yaml, pandas as pd
from rapidfuzz import fuzz
from src.common import ROOT, get_logger, ledger

log = get_logger("crossref")
UA = "WP37-UQO-research/0.1 (mailto:simon-pierre.boucher@uqo.ca)"
LIT = ROOT / "literature"

def norm(s): return re.sub(r"[^a-z0-9 ]", " ", (s or "").lower())

def query(author, title, year):
    params = {"query.bibliographic": f"{author} {title} {year}", "rows": 5, "select": "DOI,title,author,issued,container-title,volume,issue,page,type,publisher"}
    r = requests.get("https://api.crossref.org/works", params=params, headers={"User-Agent": UA}, timeout=60)
    r.raise_for_status()
    return r.json()["message"]["items"]

def main():
    seeds = yaml.safe_load((LIT / "seed_references.yaml").read_text())
    rows = []
    for s in seeds:
        if not s["author"]:
            rows.append({**s, "verified": False, "note": "placeholder topic query, needs a specific paper"}); continue
        try:
            items = query(s["author"], s["title"], s["year"])
        except Exception as e:
            rows.append({**s, "verified": False, "note": f"crossref error {type(e).__name__}"}); continue
        best, best_sim = None, 0
        for it in items:
            t = " ".join(it.get("title") or [])
            sim = fuzz.token_set_ratio(norm(t), norm(s["title"])) / 100
            yr = (it.get("issued", {}).get("date-parts") or [[None]])[0][0]
            fam = " ".join(a.get("family", "") for a in it.get("author", []) or [])
            auth_ok = all(norm(w) in norm(fam) for w in s["author"].split()[:1])
            if sim > best_sim and yr and abs(int(yr) - int(s["year"])) <= 1 and auth_ok:
                best, best_sim = it, sim
        if best and best_sim >= 0.90:
            yr = best["issued"]["date-parts"][0][0]
            rows.append({**s, "verified": True, "doi": best["DOI"], "crossref_title": " ".join(best.get("title") or []),
                         "journal": " ".join(best.get("container-title") or []), "volume": best.get("volume"), "issue": best.get("issue"),
                         "pages": best.get("page"), "pub_year": yr, "type": best.get("type"), "publisher": best.get("publisher"),
                         "authors": "; ".join(f"{a.get('family','')}, {a.get('given','')}" for a in best.get("author", []) or []),
                         "title_sim": round(best_sim, 3), "note": ""})
        else:
            rows.append({**s, "verified": False, "note": f"no confident Crossref match (best sim {best_sim:.2f})"})
        time.sleep(0.3)
    df = pd.DataFrame(rows)
    df.to_csv(LIT / "references_verified.csv", index=False)
    df[~df.verified].to_csv(LIT / "unverified.csv", index=False)
    log.info("verified %d / %d ; unverified: %s", df.verified.sum(), len(df), df.loc[~df.verified, "key"].tolist())
    ledger("2", "verify_refs", f"{int(df.verified.sum())}/{len(df)} verified via Crossref")

if __name__ == "__main__":
    main()