app/core/names.py

from __future__ import annotations

from collections import defaultdict
import re
import unicodedata

from rapidfuzz import fuzz

from app.models import FinishedMatch, TeamIdentity


ALIASES = {
    "manchester city": "man city",
    "manchester united": "man united",
    "internazionale": "inter",
    "internazionale milano": "inter",
    "inter milan": "inter",
    "paris saint germain": "psg",
    "atletico de madrid": "atletico madrid",
    "club atletico de madrid": "atletico madrid",
    "bayern munchen": "bayern munich",
    "borussia monchengladbach": "gladbach",
    "sporting clube de portugal": "sporting cp",
    "sporting lisbon": "sporting cp",
    # The Odds API expands these Dutch acronyms while football-data.org uses
    # the official short form. Keep only exact, competition-known aliases;
    # short unknown acronyms must still fail closed in fuzzy matching.
    "psv eindhoven": "psv",
    "az alkmaar": "az",
    "nec nijmegen": "nec",
}


def normalize_name(value: str) -> str:
    value = unicodedata.normalize("NFKD", value or "")
    value = "".join(ch for ch in value if not unicodedata.combining(ch))
    value = value.lower().replace("&", " and ")
    value = re.sub(r"[^a-z0-9 ]+", " ", value)
    value = re.sub(
        r"\b(fc|afc|cf|sc|ac|calcio|club|football|futebol|deportivo|fk|sv|vfl|ssc|ss|as)\b",
        " ",
        value,
    )
    value = re.sub(r"\s+", " ", value).strip()
    return ALIASES.get(value, value)


def similarity(a: str, b: str) -> float:
    na, nb = normalize_name(a), normalize_name(b)
    if not na or not nb:
        return 0.0
    if na == nb:
        return 100.0
    if min(len(na), len(nb)) <= 3:
        return 0.0
    return max(float(fuzz.WRatio(na, nb)), float(fuzz.token_set_ratio(na, nb)))


def build_team_catalog(matches: list[FinishedMatch], competition: str) -> list[TeamIdentity]:
    aliases: dict[str, set[str]] = defaultdict(set)
    names: dict[str, str] = {}

    for m in matches:
        if m.competition != competition:
            continue
        for key, name, extra in (
            (m.home_key, m.home, m.home_aliases),
            (m.away_key, m.away, m.away_aliases),
        ):
            names.setdefault(key, name)
            aliases[key].add(name)
            aliases[key].update(a for a in extra if a)

    return [
        TeamIdentity(key=key, name=names[key], aliases=tuple(sorted(aliases[key])))
        for key in sorted(names)
    ]


def _identity_score(query: str, identity: TeamIdentity) -> float:
    scores = [similarity(query, identity.name)]
    qn = normalize_name(query)
    for alias in identity.aliases:
        an = normalize_name(alias)
        if qn and qn == an:
            return 100.0
        if len(an) >= 4:
            scores.append(similarity(query, alias))
    return max(scores) if scores else 0.0


def resolve_identity(
    name: str,
    catalog: list[TeamIdentity],
    minimum: float = 82.0,
    minimum_margin: float = 6.0,
) -> tuple[TeamIdentity | None, float, float]:
    if not catalog:
        return None, 0.0, 0.0

    ranked = sorted(
        ((_identity_score(name, identity), identity) for identity in catalog),
        key=lambda item: item[0],
        reverse=True,
    )
    best_score, best = ranked[0]
    second_score = ranked[1][0] if len(ranked) > 1 else 0.0
    margin = best_score - second_score

    if best_score < minimum:
        return None, best_score, margin
    # Exact/near-exact matches are safe even if two clubs have similar long names.
    if best_score < 97.0 and margin < minimum_margin:
        return None, best_score, margin
    return best, best_score, margin


def resolve_event_pair(
    home_name: str,
    away_name: str,
    catalog: list[TeamIdentity],
    minimum: float = 82.0,
) -> tuple[TeamIdentity | None, TeamIdentity | None, float, dict[str, float]]:
    home, hs, hm = resolve_identity(home_name, catalog, minimum=minimum)
    away, as_, am = resolve_identity(away_name, catalog, minimum=minimum)

    detail = {
        "home_score": hs,
        "away_score": as_,
        "home_margin": hm,
        "away_margin": am,
    }
    if not home or not away or home.key == away.key:
        return None, None, 0.0, detail

    confidence = min(hs, as_) / 100.0
    if min(hm, am) < 8.0 and min(hs, as_) < 97.0:
        confidence *= 0.94
    return home, away, confidence, detail