"""从已落库脚本本地拆出角色/场景 —— 不调模型、不扣积分。

脚本生成时已要求结构化 entities;这里负责:
1. 读 ScriptVersion.metadata / content 里的 entities
2. 只保留 character / scene(商品不抽)
3. 至少保证 1 角色 + 1 场景(缺则按分镜补默认)
4. 回填 project.metadata + 每镜 entity_refs + entities_extracted
"""
from __future__ import annotations

import json
import re
from typing import Any


def _as_list(value: Any) -> list:
    return value if isinstance(value, list) else []


def _load_draft_entities(script) -> list[dict]:
    meta = script.metadata if isinstance(script.metadata, dict) else {}
    ents = _as_list(meta.get("entities"))
    if ents:
        return [e for e in ents if isinstance(e, dict)]
    raw = (script.content or "").strip()
    if not raw:
        return []
    try:
        data = json.loads(raw)
    except (TypeError, ValueError, json.JSONDecodeError):
        return []
    if not isinstance(data, dict):
        return []
    return [e for e in _as_list(data.get("entities")) if isinstance(e, dict)]


def _normalize_entities(raw: list[dict]) -> list[dict]:
    out: list[dict] = []
    seen: set[str] = set()
    for i, ent in enumerate(raw):
        etype = str(ent.get("type") or "").strip().lower()
        if etype in {"person", "cast", "role"}:
            etype = "character"
        if etype in {"location", "bg", "background"}:
            etype = "scene"
        if etype not in {"character", "scene"}:
            continue
        eid = str(ent.get("id") or f"{'c' if etype == 'character' else 's'}{i + 1}").strip()
        while eid in seen:
            eid = f"{eid}_{i}"
        seen.add(eid)
        name = str(ent.get("name") or eid).strip() or eid
        visual = str(ent.get("visual_prompt") or ent.get("prompt") or "").strip()
        if not visual:
            visual = (
                f"{name},全身出镜,自然光,9:16竖屏"
                if etype == "character"
                else f"{name},环境空镜,干净构图,16:9横屏"
            )
        out.append(
            {
                "id": eid,
                "type": etype,
                "name": name,
                "visual_prompt": visual,
                "ref_index": len(out) + 1,
            }
        )
    return out


def _default_character(segments) -> dict:
    # 优先用第一镜画面里像「人」的描述,否则通用主角
    hint = ""
    for seg in segments:
        text = f"{getattr(seg, 'visual_prompt', '') or ''} {getattr(seg, 'narration', '') or ''}"
        if re.search(r"女|男|人|主播|模特|客户|宝妈|姐姐|哥哥", text):
            hint = (getattr(seg, "visual_prompt", None) or text).strip()
            break
    visual = hint[:80] if hint else "电商短视频出镜主角,全身,自然妆容,干净背景,9:16竖屏"
    return {
        "id": "c1",
        "type": "character",
        "name": "主角",
        "visual_prompt": visual,
        "ref_index": 1,
    }


def _default_scene(segments) -> dict:
    hint = ""
    for seg in segments:
        text = (getattr(seg, "visual_prompt", None) or "").strip()
        if text:
            hint = text
            break
    visual = hint[:80] if hint else "干净室内场景,柔和光线,适合电商口播,16:9横屏"
    return {
        "id": "s1",
        "type": "scene",
        "name": "主场景",
        "visual_prompt": visual,
        "ref_index": 1,
    }


def _ensure_min_entities(entities: list[dict], segments) -> list[dict]:
    has_char = any(e["type"] == "character" for e in entities)
    has_scene = any(e["type"] == "scene" for e in entities)
    if not has_char:
        entities.append(_default_character(segments))
    if not has_scene:
        entities.append(_default_scene(segments))
    # 重排 ref_index
    for i, e in enumerate(entities):
        e["ref_index"] = i + 1
    return entities


def _backfill_segment_refs(segments, entities: list[dict]) -> None:
    valid = {e["id"] for e in entities}
    char_ids = [e["id"] for e in entities if e["type"] == "character"]
    scene_ids = [e["id"] for e in entities if e["type"] == "scene"]
    default_refs = (char_ids[:1] + scene_ids[:1]) or list(valid)[:2]
    for seg in segments:
        refs = [r for r in (seg.entity_refs or []) if r in valid]
        speaker = (seg.speaker or "").strip()
        if speaker and speaker in valid and speaker not in refs:
            refs.append(speaker)
        if not refs:
            refs = list(default_refs)
        if refs != (seg.entity_refs or []):
            seg.entity_refs = refs
            seg.save(update_fields=["entity_refs", "updated_at"])


def materialize_script_entities(*, project, script) -> list[dict]:
    """本地拆角色/场景并落库。返回规范化后的 entities 列表。"""
    from django.db import transaction

    from apps.ai.script_agent import _map_entities_to_project_metadata

    segments = list(script.segments.order_by("sort_order"))
    entities = _ensure_min_entities(_normalize_entities(_load_draft_entities(script)), segments)
    with transaction.atomic():
        _map_entities_to_project_metadata(project, entities)
        md = dict(project.metadata or {})
        md["entities_extracted"] = True
        md["entities_extract_mode"] = "local"
        project.metadata = md
        project.save(update_fields=["metadata", "updated_at"])
        _backfill_segment_refs(segments, entities)
        # 同步写回脚本 metadata.entities,方便下次读
        sm = dict(script.metadata or {})
        sm["entities"] = entities
        script.metadata = sm
        script.save(update_fields=["metadata", "updated_at"])
    return entities