# -*- coding: utf-8 -*-
"""引用完整性检查 v2：全角标点 + 分块N解析 + KB内部引用"""
import os, re

BASE = r"D:/AgentSkill/MCNVideo AI/project/短视频脚本创作/V1.0/脚本创作技能"
KB = os.path.join(BASE, "references", "知识库")

kb_files = {}
for d in sorted(os.listdir(KB)):
    dp = os.path.join(KB, d)
    if os.path.isdir(dp):
        kb_files[d] = sorted(f for f in os.listdir(dp) if f.endswith(".md"))
total = sum(len(v) for v in kb_files.values())

# 分块前缀分类：文件名以"分块N_"开头的分类，允许 分块N 引用方式
block_cats = {c: {int(re.match(r'分块(\d+)_', f).group(1)): f for f in fs if re.match(r'分块(\d+)_', f)}
              for c, fs in kb_files.items()}

scan_files = []
for sub in ["创作流程"]:
    p = os.path.join(BASE, "references", sub)
    for f in sorted(os.listdir(p)):
        if f.endswith(".md"):
            scan_files.append(("流程/" + f, os.path.join(p, f)))
scan_files.append(("创作流程规范.md", os.path.join(BASE, "references", "创作流程规范.md")))
scan_files.append(("SKILL.md", os.path.join(BASE, "SKILL.md")))
# 知识库内部交叉引用
for c, fs in kb_files.items():
    for f in fs:
        scan_files.append((f"知识库/{c}/{f}", os.path.join(KB, c, f)))

FN = r'[\w（）()·：:\-\\\u4e00-\u9fff/]+?\.md'
file_ref_re = re.compile(FN)
cat_name_re = re.compile(r'(\d{2}_[\w（）()·\-\u4e00-\u9fff]+)')
block_re = re.compile(r'分块(\d+)')
cat_set = set(kb_files.keys())

referenced = set()   # (cat, fn)
broken = []
for label, path in scan_files:
    try:
        lines = open(path, encoding="utf-8").read().splitlines()
    except Exception as e:
        print(f"READ FAIL {label}: {e}")
        continue
    for i, line in enumerate(lines, 1):
        cats_in_line = [c for c in cat_set if c in line]
        # 文件名引用
        for m in file_ref_re.finditer(line):
            ref = m.group(0).replace("\\", "/")
            fn = ref.split("/")[-1]
            # 尝试带路径匹配
            hit = None
            for c, fs in kb_files.items():
                if fn in fs:
                    hit = (c, fn)
            if hit:
                referenced.add(hit)
            elif "知识库" in line or ref.startswith("知识库") or "/" in ref:
                # 可能是知识库引用但没匹配上（也排除产物/流程自身引用）
                if not any(fn == x for x in ["创作流程规范.md"]) and not fn.startswith(("0", "1")) or "知识库" in ref:
                    if "知识库" in ref or ("知识库" in line and ("/" in ref)):
                        broken.append((label, i, ref, line.strip()[:100]))
        # 分块N引用（限当前行有分类上下文）
        for m in block_re.finditer(line):
            n = int(m.group(1))
            for c in cats_in_line:
                if c in block_cats and n in block_cats[c]:
                    referenced.add((c, block_cats[c][n]))

print(f"总文件: {total}")
print("\n===== A. 疑似断链（引用了知识库路径但未命中文件） =====")
seen = set()
for b in broken:
    k = (b[2],)
    if k in seen: continue
    seen.add(k)
    print(f"  {b[0]}:{b[1]} -> {b[2]}\n     {b[3]}")
if not broken:
    print("  无")

print("\n===== B. 零引用文件（流程+规范+SKILL+知识库内部均未引用文件名/分块号） =====")
orphans = []
for c, fs in kb_files.items():
    for fn in fs:
        if (c, fn) not in referenced:
            orphans.append(f"{c}/{fn}")
print(f"零引用: {len(orphans)} / {total}")
for o in orphans:
    print(f"  {o}")

print("\n===== C. 分类引用统计 =====")
for c, fs in kb_files.items():
    used = sum(1 for fn in fs if (c, fn) in referenced)
    print(f"  {c}: {used}/{len(fs)}")