mtg/tools/cards.py

#!/usr/bin/env python3
"""Shared card index: joins Scryfall oracle + Tagger tags + MTGJSON edhrecRank
+ the Commander Spellbook combo DB into one in-memory object.

Every other tool builds on this.

    from cards import DB
    db = DB()
    db["Necroskitter"].tags              -> {'death-trigger-opponent', ...}
    db.tagged("free-sacrifice-outlet")   -> [Card, ...]
    db.identity_legal(card, identity)    -> bool

Color identity is never hardcoded here -- it's always passed in by the caller,
derived from whichever commander that deck's config.yaml names (see
tools/deckctx.py DeckContext.identity()).
"""
import json, os, functools, re, zipfile
from dataclasses import dataclass, field

DATA = os.path.join(os.path.dirname(__file__), "..", "data")


@dataclass(eq=False)   # identity hash: cards are unique by object
class Card:
    name: str
    mana_cost: str = ""
    mana_value: float = 0.0
    type_line: str = ""
    oracle: str = ""
    color_identity: frozenset = frozenset()
    legal_commander: bool = False
    game_changer: bool = False
    edhrec_rank: int | None = None
    tags: set = field(default_factory=set)
    # Forge card-script facts (see DB._load_forge_scripts)
    ai_removed: bool = False        # AI:RemoveDeck:All -- Forge will NEVER play it
    ai_logic: str | None = None     # AILogic$ <name> -- has a bespoke AI routine
    ai_sac_pref: bool = False       # SVar:AIPreference:SacCost -- knows what to sac

    @property
    def is_land(self):
        return "Land" in self.type_line

    def __repr__(self):
        return f"<{self.name} {self.mana_cost}>"


class DB:
    def __init__(self, data_dir=DATA, blacklist_path=None):
        self.dir = data_dir
        self.cards: dict[str, Card] = {}
        self._by_tag: dict[str, list] = {}
        self._load_oracle()
        self._load_tags()
        self._load_edhrec()
        # blacklist is per-deck (a deck's own preferences), not shared data --
        # callers normally pass DeckContext.blacklist_path. With no path given
        # this loads nothing, which is a safe default for deck-agnostic tools.
        self.blacklist = load_blacklist(blacklist_path) if blacklist_path else {}
        self._load_forge_scripts()

    # ---------- loading ----------
    def _load_oracle(self):
        for line in open(f"{self.dir}/oracle-cards.jsonl"):
            c = json.loads(line)
            if c.get("layout") in ("token", "double_faced_token", "art_series", "emblem"):
                continue
            faces = c.get("card_faces") or []
            self.cards[c["name"]] = Card(
                name=c["name"],
                mana_cost=c.get("mana_cost") or " // ".join(f.get("mana_cost", "") for f in faces),
                mana_value=c.get("cmc", 0.0),
                type_line=c.get("type_line", ""),
                oracle=c.get("oracle_text") or "\n".join(f.get("oracle_text", "") for f in faces),
                color_identity=frozenset(c.get("color_identity", [])),
                legal_commander=c.get("legalities", {}).get("commander") == "legal",
                game_changer=bool(c.get("game_changer")),
            )
            self.cards[c["name"]]._oracle_id = c.get("oracle_id")

    def _load_tags(self):
        by_oid = {}
        for line in open(f"{self.dir}/oracle-tags.jsonl"):
            t = json.loads(line)
            for tg in t["taggings"]:
                by_oid.setdefault(tg["oracle_id"], set()).add(t["slug"])
        for card in self.cards.values():
            card.tags = by_oid.get(getattr(card, "_oracle_id", None), set())
            for slug in card.tags:
                self._by_tag.setdefault(slug, []).append(card)

    def _load_edhrec(self):
        atomic = json.load(open(f"{self.dir}/AtomicCards.json"))["data"]
        for name, printings in atomic.items():
            c = self.cards.get(name)
            if c and printings:
                c.edhrec_rank = printings[0].get("edhrecRank")

    def _load_forge_scripts(self, forge=None):
        """Read Forge's card scripts to learn which cards its AI can actually play.

        `AI:RemoveDeck:All` means Forge is forbidden from ever playing the card --
        2,525 cards carry it. Those show up as "never cast" in sim results, which
        is a SIMULATOR ARTIFACT, not evidence the card is bad. Distinguishing the
        two is the whole point of this method.
        """
        forge = forge or os.path.expanduser("~/forge")
        zpath = os.path.join(forge, "res", "cardsfolder", "cardsfolder.zip")
        if not os.path.exists(zpath):
            return
        cache = os.path.join(self.dir, ".forge_scripts.json")
        if os.path.exists(cache) and os.path.getmtime(cache) > os.path.getmtime(zpath):
            info = json.load(open(cache))
        else:
            info = {}
            with zipfile.ZipFile(zpath) as z:
                for n in z.namelist():
                    if not n.endswith(".txt"):
                        continue
                    try:
                        txt = z.read(n).decode("utf-8", "ignore")
                    except Exception:
                        continue
                    first = txt.split("\n", 1)[0]
                    if not first.startswith("Name:"):
                        continue
                    name = first[5:].strip()
                    m = re.search(r"AILogic\$\s*([A-Za-z.]+)", txt)
                    info[name] = {
                        "ai_removed": "AI:RemoveDeck:All" in txt,
                        "ai_logic": m.group(1) if m else None,
                        "ai_sac_pref": "AIPreference:SacCost" in txt,
                    }
            try:
                json.dump(info, open(cache, "w"))
            except OSError:
                pass
        for name, d in info.items():
            c = self.cards.get(name)
            if c:
                c.ai_removed = d["ai_removed"]
                c.ai_logic = d["ai_logic"]
                c.ai_sac_pref = d["ai_sac_pref"]

    # ---------- queries ----------
    def __getitem__(self, name): return self.cards[name]
    def get(self, name):         return self.cards.get(name)

    def tagged(self, *slugs, identity=None, legal=True, allow_blacklisted=False):
        """Cards carrying ALL given tags, within a color identity.

        `identity` is a frozenset of color letters (e.g. frozenset("BRG")),
        normally from DeckContext.identity() -- never a hardcoded combination.
        identity=None means no color filter at all.
        Blacklisted cards are omitted unless allow_blacklisted=True.
        """
        out = None
        for s in slugs:
            hits = set(self._by_tag.get(s, []))
            out = hits if out is None else out & hits
        out = out or set()
        res = [c for c in out
               if (identity is None or c.color_identity <= identity)
               and (not legal or c.legal_commander)
               and (allow_blacklisted or c.name not in self.blacklist)]
        return sorted(res, key=lambda c: c.edhrec_rank or 10**9)

    def identity_legal(self, c, identity):
        """Is this card legal in Commander and within the given color identity?"""
        return c.legal_commander and c.color_identity <= identity

    @functools.cached_property
    def combos(self):
        """Commander Spellbook variants, indexed by card name."""
        v = json.load(open(f"{self.dir}/variants.json"))["variants"]
        idx = {}
        for combo in v:
            for u in combo["uses"]:
                idx.setdefault(u["card"]["name"], []).append(combo)
        return v, idx


def load_blacklist(path="blacklist.txt"):
    """Cards excluded by personal preference. Returns {name: reason}.

    Distinct from legality: these are choices, not rules. Kept in one file so
    both the checker and the search tools honour them.
    """
    out = {}
    if not os.path.exists(path):
        return out
    for raw in open(path):
        line = raw.strip()
        if not line or line.startswith("#"):
            continue
        name, _, reason = line.partition("|")
        out[name.strip()] = reason.strip() or "no reason given"
    return out


def load_decklist(path):
    """Parse a decklist. '# Commander' / '# Main' headers select the section.
    Lines are 'N Card Name' or just 'Card Name'. Returns (commander, [names])."""
    commander, main, section = None, [], "main"
    for raw in open(path):
        line = raw.strip()
        if not line:
            continue
        if line.startswith("#"):
            # only top-level '# ' headers switch sections; '## ' are just comments
            if not line.startswith("##"):
                low = line.lower()
                if "commander" in low:
                    section = "commander"
                elif "main" in low or "deck" in low:
                    section = "main"
            continue
        parts = line.split(None, 1)
        if parts[0].isdigit() and len(parts) > 1:
            n, name = int(parts[0]), parts[1].strip()
        else:
            n, name = 1, line
        if section == "commander":
            commander = name
        else:
            main.extend([name] * n)
    return commander, main


if __name__ == "__main__":
    db = DB()
    print(f"{len(db.cards)} cards, {len(db._by_tag)} tags")
    print("gamechangers:", sum(1 for c in db.cards.values() if c.game_changer))
    for c in db.tagged("death-trigger-opponent")[:5]:
        print(" ", c.edhrec_rank, c.name)