#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""Structural evidence and diagram QA gate. Humans must still verify real physics."""
import argparse,json,re,sys,xml.etree.ElementTree as ET
from pathlib import Path
ROOT=Path(__file__).resolve().parents[1]

def run(strict=True):
    d=json.loads((ROOT/'story.json').read_text(encoding='utf-8'))
    errors=[];warnings=[]
    cl=d.get('claims',[])
    if not cl:errors.append('缺少科学主张及对应证据 claims')
    ids=set()
    for c in cl:
        key=c.get('id')
        if not key or key in ids:errors.append('科学主张 ID 不唯一或为空')
        ids.add(key)
        if not c.get('statement'):errors.append(f'{key}: 主张内容不能为空')
        if not c.get('sources') or any(not re.match(r'^https?://',x) for x in c.get('sources',[])):
            errors.append(f'{key}: 至少一条可核实的 HTTP(S) 来源')
        if not c.get('diagram_truth'):errors.append(f'{key}: 必须列明示意图不容颠倒的科学关系')
        if c.get('source_checked') is not True:
            warnings.append(f'{key}: 来源尚未由制作者人工/浏览器核对；不得标记已审稿')
    covered=set()
    scenes=d.get('scenes',[])
    if len(scenes)<3:errors.append('科普视频至少 3 幕：现象、原理、总结')
    for i,sc in enumerate(scenes):
        sceneid=sc.get('id',f'#{i}')
        links=sc.get('claim_ids',[])
        if not links:errors.append(f'{sceneid}: 缺少 claim_ids')
        for link in links:
            if link not in ids:errors.append(f'{sceneid}: 未知科学主张 {link}')
            covered.add(link)
        art=sc.get('diagram_svg') or sc.get('image')
        if not art:
            errors.append(f'{sceneid}: 缺少 diagram_svg/image，不能仅凭标题渲染')
        elif sc.get('diagram_svg'):
            loc=(ROOT/sc['diagram_svg']).resolve()
            if not loc.exists():errors.append(f'{sceneid}: SVG 不存在：{sc["diagram_svg"]}')
            else:
                txt=loc.read_text(encoding='utf-8')
                if 'data-placeholder="true"' in txt:
                    errors.append(f'{sceneid}: SVG 还是占位画面，禁止发布')
                try: ET.fromstring(txt)
                except ET.ParseError: errors.append(f'{sceneid}: SVG 语法损坏')
        if not sc.get('visual_truth'):
            errors.append(f'{sceneid}: 缺少物理关系/方向/颜色/空间约束 visual_truth')
        if len(sc.get('heading',''))>22:warnings.append(f'{sceneid}: 标题过长，检查竖屏适配')
        if len(sc.get('subtitle',''))>36:warnings.append(f'{sceneid}: 说明文字过长，易与字幕重叠')
    for key in ids-covered: warnings.append(f'{key}: 尚未在分镜中展示该科学主张')
    v=d.get('voices',{})
    for role,cfg in v.items():
        if not isinstance(cfg,dict) or not cfg.get('speaker'):errors.append(f'{role}: 声音配置不完整')
    report={'science_gate_pass':not errors and (not strict or not warnings), 'hard_errors':errors,'human_review_pending':warnings,
            'scientific_truth_manually_verified':False,
            'claims':len(cl),'scenes':len(scenes),
            'note':'本脚本仅检查证据链和示意图规格，不能自动证明光路或公式正确。'}
    (ROOT/'output').mkdir(exist_ok=True)
    (ROOT/'output/science-audit.json').write_text(json.dumps(report,ensure_ascii=False,indent=2),encoding='utf-8')
    print(json.dumps(report,ensure_ascii=False,indent=2))
    return 0 if report['science_gate_pass'] else 2

if __name__=='__main__':
    p=argparse.ArgumentParser();p.add_argument('--allow-unreviewed',action='store_true',help='Only structural drafting; never count as science approval');a=p.parse_args()
    sys.exit(run(strict=not a.allow_unreviewed))