"""Tests purs : canonicalisation des noms variants + anonymes par genre/age.

`_canonicalize_speakers`, `_apply_anonymous_speakers` et `_anon_identity` sont
deterministes et testables sans Gemma ni disque (cf. test_incises.py).
"""
from __future__ import annotations

from inkflow.analysis.segmenter import (
    _anon_identity,
    _apply_anonymous_speakers,
    _canonicalize_speakers,
    _inversion_gender,
    _resolve_anonymous_figurants,
)
from inkflow.models import Character, Incise, Segment, SegmentType


def _C(name, gender=None, age=None, aliases=None):
    return Character(name=name, gender=gender, age=age, aliases=aliases or [])


def _D(text, speaker, incises=None):
    return Segment(type=SegmentType.DIALOGUE, text=text, speaker=speaker,
                   incises=incises or [])


def _N(text="narration"):
    return Segment(type=SegmentType.NARRATION, text=text, speaker="narrateur")


# --- Canonicalisation des variantes de noms ----------------------------------

def test_canon_variante_vers_canonique():
    chars = [_C("Sagale"), _C("Elvi"), _C("Holden")]
    segs = [_D("a", "Amiral Mehmet Sagale"), _D("b", "Elvi Okoye"),
            _D("c", "Holden")]
    _canonicalize_speakers(segs, chars)
    assert [s.speaker for s in segs] == ["Sagale", "Elvi", "Holden"]


def test_canon_reciproque_forme_courte_vers_complete():
    # Le cast porte le nom complet ; une surface courte distinctive s'y recolle.
    chars = [_C("Elvi Okoye")]
    segs = [_D("a", "Okoye")]
    _canonicalize_speakers(segs, chars)
    assert segs[0].speaker == "Elvi Okoye"


def test_canon_marine_unique_distinctif():
    chars = [_C("Marine"), _C("Holden")]
    segs = [_D("a", "Marine de gauche")]
    _canonicalize_speakers(segs, chars)
    assert segs[0].speaker == "Marine"


def test_canon_ambiguite_sabstient():
    # Deux personnages partagent le token "marine" -> non distinctif -> abstention.
    chars = [_C("Marine Lopez"), _C("Marine Cho")]
    segs = [_D("a", "Marine de gauche")]
    _canonicalize_speakers(segs, chars)
    assert segs[0].speaker == "Marine de gauche"  # inchange


def test_canon_inconnu_total_inchange():
    chars = [_C("Holden"), _C("Kajri")]
    segs = [_D("a", "Bob")]
    _canonicalize_speakers(segs, chars)
    assert segs[0].speaker == "Bob"


def test_canon_narrateur_et_inconnu_jamais_touches():
    chars = [_C("Sagale")]
    segs = [_N(), _D("a", "inconnu"), _D("b", "?")]
    _canonicalize_speakers(segs, chars)
    assert [s.speaker for s in segs] == ["narrateur", "inconnu", "?"]


def test_canon_idempotent():
    chars = [_C("Sagale")]
    segs = [_D("a", "Amiral Mehmet Sagale")]
    _canonicalize_speakers(segs, chars)
    once = segs[0].speaker
    _canonicalize_speakers(segs, chars)
    assert segs[0].speaker == once == "Sagale"


# --- Identite anonyme par (genre, age) ---------------------------------------

def test_anon_identity_format():
    assert _anon_identity("male", "adult") == "anonyme (homme, adulte)"
    assert _anon_identity("male", None) == "anonyme (homme)"
    assert _anon_identity("female", None) == "anonyme (femme)"
    assert _anon_identity(None, None) == "anonyme"
    assert _anon_identity(None, "child") == "anonyme (enfant)"


def test_apply_anonymous_role_par_genre():
    # "informa le soldat" -> anonyme (homme) ; renvoie le bucket avec genre/age.
    t = "La réception commence, madame, informa le soldat."
    inc = Incise(start=t.index("informa"), end=len(t))
    segs = [_D(t, "inconnu", [inc])]
    used = _apply_anonymous_speakers(segs, names={"Kajri"})
    assert segs[0].speaker == "anonyme (homme)"
    assert used == {"anonyme (homme)": ("male", None)}


def test_apply_anonymous_role_inconnu_genre():
    # "une voix" : role sans genre fiable -> bucket generique "anonyme".
    t = "Par ici, indiqua une voix."
    inc = Incise(start=t.index("indiqua"), end=len(t))
    segs = [_D(t, "inconnu", [inc])]
    used = _apply_anonymous_speakers(segs, names=set())
    assert segs[0].speaker == "anonyme"
    assert used == {"anonyme": (None, None)}


def test_apply_anonymous_ignore_nom_propre():
    # Incise a nom propre -> pas un anonyme, speaker inchange.
    t = "Bonjour, lança Drummer."
    inc = Incise(start=t.index("lança"), end=len(t))
    segs = [_D(t, "Drummer", [inc])]
    used = _apply_anonymous_speakers(segs, names={"Drummer"})
    assert segs[0].speaker == "Drummer"
    assert used == {}


# --- Rang/titre devant un nom propre -----------------------------------------

def test_rang_titre_capte_le_nom_propre():
    # "dit l'amiral Sagale" : le rang n'est pas un anonyme, on capte "Sagale".
    from inkflow.analysis.segmenter import detect_incises, incise_role, incise_speaker
    t = "Dr Okoye, dit l'amiral Sagale."
    inc = detect_incises(t, names={"Sagale"})[0]
    assert incise_speaker(t, inc, {"Sagale"}) == "Sagale"
    assert incise_role(t, inc, {"Sagale"}) is None


# --- Stabilite du nom canonique etabli ---------------------------------------

def test_reconcile_garde_nom_etabli_stable():
    # Un nom deja dans le cast ("Sagale") n'est pas renomme par une forme plus
    # longue trouvee dans un chapitre ("Amiral Mehmet Sagale") -> alias.
    from inkflow.casting.dedup import reconcile_characters
    book = [_C("Sagale", gender="male")]
    found = [_C("Amiral Mehmet Sagale", gender="male")]
    chars, _ = reconcile_characters(book, found, None)
    sagale = next(c for c in chars if c.name == "Sagale")
    assert "Amiral Mehmet Sagale" in sagale.aliases


def test_reconcile_nouveau_perso_garde_forme_complete():
    # Sans nom etabli, le comportement reste "la forme la plus complete gagne".
    from inkflow.casting.dedup import reconcile_characters
    chars, _ = reconcile_characters([], [_C("Jim"), _C("Jim Holden")], None)
    assert any(c.name == "Jim Holden" and "Jim" in c.aliases for c in chars)


# --- Figurants anonymes resolus via la narration adjacente -------------------

def test_inversion_gender():
    assert _inversion_gender("Souhaitez-vous une escorte ? demanda-t-elle.") == "female"
    assert _inversion_gender("Stop, dit-il.") == "male"
    assert _inversion_gender("Je pars maintenant.") is None


def test_figurant_femme_via_narration_avant():
    # Replique indeterminee + narration decrivant "La jeune marine" -> anonyme femme.
    segs = [
        _N("La jeune marine toucha quelque chose au poignet de son armure."),
        _D("Prévenez-nous quand vous serez prête à ressortir.", "inconnu"),
    ]
    used = _resolve_anonymous_figurants(segs)
    assert segs[1].speaker == "anonyme (femme)"
    assert "anonyme (femme)" in used


def test_figurant_genre_par_pronom_inversion_prioritaire():
    # "demanda-t-elle" (féminin) prime, narration "Le soldat" -> on garde femme.
    segs = [
        _N("Le soldat s'avança vers eux."),
        _D("Souhaitez-vous une escorte ? demanda-t-elle.", "?"),
    ]
    _resolve_anonymous_figurants(segs)
    assert segs[0].speaker == "narrateur"
    assert segs[1].speaker == "anonyme (femme)"


def test_figurant_ne_touche_pas_les_resolus():
    # Une replique deja attribuee n'est jamais ecrasee, meme avec narration de role.
    segs = [
        _N("Le soldat montait la garde."),
        _D("J'arrive.", "Holden"),
    ]
    _resolve_anonymous_figurants(segs)
    assert segs[1].speaker == "Holden"


def test_figurant_sans_narration_de_role_inchange():
    segs = [_N("La pièce était sombre."), _D("Qui est là ?", "inconnu")]
    _resolve_anonymous_figurants(segs)
    assert segs[1].speaker == "inconnu"