AntiCoco/hellofresh/filter.py

"""Filtrage des recettes : exclusion d'ingrédients (coco !) + scoring par préférences.

Matching insensible à la casse ET aux accents : « Noix de Coco », « noix de coco rapée »
et « creme de coco » matchent tous l'entrée « coco ». On compare des mots normalisés sur
le nom des ingrédients, les allergènes, le nom et le titre de la recette.
"""

from __future__ import annotations

import json
import unicodedata
from pathlib import Path

from .auth import ROOT
from .models import Recipe

EXCLUDES_PATH = ROOT / "config" / "excludes.json"
PREFS_PATH = ROOT / "config" / "prefs.json"


def normalize(s: str) -> str:
    """Minuscule + suppression des accents (NFD → drop des diacritiques)."""
    s = unicodedata.normalize("NFD", s or "")
    s = "".join(c for c in s if unicodedata.category(c) != "Mn")
    return s.lower().strip()


# --- gestion de la liste d'exclusion ---------------------------------------
def load_excludes() -> list[str]:
    if not EXCLUDES_PATH.exists():
        return []
    data = json.loads(EXCLUDES_PATH.read_text(encoding="utf-8"))
    return list(data.get("exclude", []))


def save_excludes(terms: list[str]) -> None:
    existing = {}
    if EXCLUDES_PATH.exists():
        existing = json.loads(EXCLUDES_PATH.read_text(encoding="utf-8"))
    existing["exclude"] = terms
    EXCLUDES_PATH.write_text(json.dumps(existing, indent=2, ensure_ascii=False) + "\n", encoding="utf-8")


def add_exclude(term: str) -> list[str]:
    terms = load_excludes()
    if normalize(term) not in {normalize(t) for t in terms}:
        terms.append(term)
        save_excludes(terms)
    return terms


def remove_exclude(term: str) -> list[str]:
    nt = normalize(term)
    terms = [t for t in load_excludes() if normalize(t) != nt]
    save_excludes(terms)
    return terms


def load_prefs() -> dict:
    if not PREFS_PATH.exists():
        return {"liked": [], "disliked": [], "allow_premium": False}
    data = json.loads(PREFS_PATH.read_text(encoding="utf-8"))
    return {
        "liked": data.get("liked", []),
        "disliked": data.get("disliked", []),
        "allow_premium": bool(data.get("allow_premium", False)),
    }


def save_prefs(liked: list[str], disliked: list[str]) -> dict:
    """Écrit liked/disliked en préservant le reste du fichier (commentaire, allow_premium)."""
    existing = {}
    if PREFS_PATH.exists():
        existing = json.loads(PREFS_PATH.read_text(encoding="utf-8"))
    existing["liked"] = liked
    existing["disliked"] = disliked
    PREFS_PATH.write_text(json.dumps(existing, indent=2, ensure_ascii=False) + "\n", encoding="utf-8")
    return {"liked": liked, "disliked": disliked}


def load_allow_premium() -> bool:
    """Réglage : True si les recettes à supplément (premium) sont autorisées par défaut."""
    return load_prefs()["allow_premium"]


def save_allow_premium(value: bool) -> bool:
    """Persiste le réglage `allow_premium` en préservant le reste de prefs.json."""
    existing = {}
    if PREFS_PATH.exists():
        existing = json.loads(PREFS_PATH.read_text(encoding="utf-8"))
    existing["allow_premium"] = bool(value)
    PREFS_PATH.write_text(json.dumps(existing, indent=2, ensure_ascii=False) + "\n", encoding="utf-8")
    return bool(value)


# --- application aux recettes ----------------------------------------------
def _exclusion_haystack(recipe: Recipe) -> str:
    """Champs FAISANT FOI pour l'exclusion : ingrédients, allergènes, nom, accroche.

    On EXCLUT volontairement les `tags` : HelloFresh y pose des tags internes non
    affichés (ex. un tag `coconut` présent sur des dizaines de recettes sans coco),
    ce qui provoquait des faux positifs massifs.
    """
    parts = [recipe.name, recipe.headline, *recipe.ingredients, *recipe.allergens]
    return normalize(" | ".join(p for p in parts if p))


def _pref_haystack(recipe: Recipe) -> str:
    """Pour le scoring de préférences, les tags sont utiles (catégories de cuisine…)."""
    parts = [recipe.name, recipe.headline, *recipe.ingredients, *recipe.tags]
    return normalize(" | ".join(p for p in parts if p))


def mark_excluded(recipe: Recipe, excludes: list[str] | None = None) -> Recipe:
    """Remplit `contains_excluded` et `matched_excludes` sur la recette."""
    excludes = excludes if excludes is not None else load_excludes()
    hay = _exclusion_haystack(recipe)
    matched = [term for term in excludes if normalize(term) and normalize(term) in hay]
    recipe.matched_excludes = matched
    recipe.contains_excluded = bool(matched)
    return recipe


def score(recipe: Recipe, prefs: dict | None = None) -> float:
    prefs = prefs or load_prefs()
    hay = _pref_haystack(recipe)
    s = 0.0
    for kw in prefs.get("liked", []):
        if normalize(kw) and normalize(kw) in hay:
            s += 1.0
    for kw in prefs.get("disliked", []):
        if normalize(kw) and normalize(kw) in hay:
            s -= 1.0
    recipe.score = s
    return s


def annotate(recipes: list[Recipe]) -> list[Recipe]:
    """Marque exclusions + score sur une liste de recettes (in place)."""
    excludes = load_excludes()
    prefs = load_prefs()
    for r in recipes:
        mark_excluded(r, excludes)
        score(r, prefs)
    return recipes


def propose(recipes: list[Recipe], count: int | None = None,
            allow_premium: bool = False) -> list[Recipe]:
    """Retire les recettes exclues (coco…) et payantes, classe le reste par score décroissant.

    `allow_premium=True` conserve les recettes à supplément (hors abonnement) dans la liste.
    """
    annotate(recipes)
    safe = [r for r in recipes
            if not r.contains_excluded and (allow_premium or not r.is_premium)]
    safe.sort(key=lambda r: r.score, reverse=True)
    return safe[:count] if count else safe