#!/usr/bin/env python3
"""
知识点覆盖率分析 + 回填脚本
用法: python3 knowledge_coverage_analyzer.py <飞书文档ID或URL>

功能:
1. 解析剧本主表，区分 TL 行（输入）与组件行（输出）
2. 读取「知识点现状」表格中的单词和句型
3. 统计每个知识点在输入/输出中的出现次数
4. 回填到「知识点现状」表格
5. 输出用法质量检查报告

依赖: lark-cli（用于读取文档 markdown），Python3（用于 API 回填）
"""

import json
import re
import subprocess
import sys
import time
import urllib.request
from pathlib import Path

# ── 配置 ──────────────────────────────────────────────
CONFIG_PATH = Path("/root/.openclaw/credentials/xiaoyan/config.json")
LARK_CLI_ENV = {
    "LARKSUITE_CLI_CONFIG_DIR": "/root/.openclaw/credentials/xiaoyan",
    "PATH": "/root/.nvm/versions/node/v24.14.0/bin:" + "/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
}


# ── 工具函数 ──────────────────────────────────────────

def parse_doc_id(url_or_id):
    """从 URL 或直接 ID 中提取文档 ID"""
    m = re.search(r"docx/([A-Za-z0-9]+)", url_or_id)
    if m:
        return m.group(1)
    return url_or_id.strip()


def fetch_markdown(doc_id):
    """用 lark-cli 获取文档 markdown"""
    result = subprocess.run(
        ["/root/.nvm/versions/node/v24.14.0/bin/lark-cli", "docs", "+fetch", "--doc", doc_id, "--as", "bot"],
        capture_output=True, text=True, env=LARK_CLI_ENV, timeout=30,
    )
    data = json.loads(result.stdout)
    return data["data"]["markdown"]


def get_token():
    """获取 tenant_access_token"""
    with open(CONFIG_PATH) as f:
        cfg = json.load(f)
    app = cfg["apps"][0]
    body = json.dumps({"app_id": app["appId"], "app_secret": app["appSecret"]}).encode()
    req = urllib.request.Request(
        "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal",
        data=body,
        headers={"Content-Type": "application/json"},
    )
    resp = json.loads(urllib.request.urlopen(req).read())
    return resp["tenant_access_token"]


def api_get(token, path):
    """GET 请求飞书 API"""
    req = urllib.request.Request(
        f"https://open.feishu.cn/open-apis{path}",
        headers={"Authorization": f"Bearer {token}"},
    )
    return json.loads(urllib.request.urlopen(req).read())


def api_patch(token, path, body):
    """PATCH 请求飞书 API"""
    data = json.dumps(body).encode()
    req = urllib.request.Request(
        f"https://open.feishu.cn/open-apis{path}",
        data=data,
        headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"},
        method="PATCH",
    )
    return json.loads(urllib.request.urlopen(req).read())


def api_get_all_children(token, doc_id, block_id):
    """分页获取 block 的所有子块"""
    all_items = []
    page_token = None
    while True:
        path = f"/docx/v1/documents/{doc_id}/blocks/{block_id}/children?page_size=50"
        if page_token:
            path += f"&page_token={page_token}"
        resp = api_get(token, path)
        items = resp.get("data", {}).get("items", [])
        all_items.extend(items)
        if not resp.get("data", {}).get("has_more"):
            break
        page_token = resp["data"]["page_token"]
    return all_items


# ── Markdown 解析 ─────────────────────────────────────

def parse_markdown_table(md, heading_text):
    """找到标题下方的 lark-table，返回二维数组 [row][col]"""
    idx = md.find(heading_text)
    if idx == -1:
        return None

    table_match = re.search(r"<lark-table.*?</lark-table>", md[idx:], re.DOTALL)
    if not table_match:
        return None

    table_md = table_match.group(0)
    rows = re.findall(r"<lark-tr>(.*?)</lark-tr>", table_md, re.DOTALL)

    result = []
    for row in rows:
        cells = re.findall(r"<lark-td[^>]*>(.*?)</lark-td>", row, re.DOTALL)
        clean_cells = []
        for c in cells:
            # 去掉所有 HTML 标签，提取纯文本
            text = re.sub(r"<[^>]+>|\*\*|\{[^}]*\}", "", c).strip()
            clean_cells.append(text)
        result.append(clean_cells)

    return result


def find_table_by_first_cell_md(md, first_cell_text):
    """在 markdown 中找到第一列第一行为指定文本的表格"""
    tables = re.findall(r"<lark-table.*?</lark-table>", md, re.DOTALL)
    for table_md in tables:
        rows = re.findall(r"<lark-tr>(.*?)</lark-tr>", table_md, re.DOTALL)
        if not rows:
            continue
        cells = re.findall(r"<lark-td[^>]*>(.*?)</lark-td>", rows[0], re.DOTALL)
        if cells:
            text = re.sub(r"<[^>]+>|\*\*|\{[^}]*\}", "", cells[0]).strip()
            if text == first_cell_text:
                return parse_markdown_table_from_string(table_md)
    return None


def parse_markdown_table_from_string(table_md):
    """从 lark-table markdown 字符串解析为二维数组"""
    rows = re.findall(r"<lark-tr>(.*?)</lark-tr>", table_md, re.DOTALL)
    result = []
    for row in rows:
        cells = re.findall(r"<lark-td[^>]*>(.*?)</lark-td>", row, re.DOTALL)
        clean_cells = [re.sub(r"<[^>]+>|\*\*|\{[^}]*\}", "", c).strip() for c in cells]
        result.append(clean_cells)
    return result


# ── 统计逻辑 ──────────────────────────────────────────

def count_occurrences(word, texts):
    """统计单词在文本列表中的出现次数（不区分大小写，整词匹配）"""
    if not word:
        return 0
    count = 0
    pattern = re.compile(r"\b" + re.escape(word) + r"\b", re.IGNORECASE)
    for t in texts:
        count += len(pattern.findall(t))
    return count


def count_pattern_occurrences(pattern_text, texts):
    """统计句型在文本列表中的出现次数（子串匹配，不区分大小写）"""
    if not pattern_text:
        return 0
    key = pattern_text.strip().rstrip(".").rstrip("!").rstrip("?")
    key_lower = key.lower()
    count = 0
    for t in texts:
        count += t.lower().count(key_lower)
    return count


# ── 回填逻辑（使用 Block API）─────────────────────────

def find_table_block_id(token, doc_id, heading_text):
    """找到标题下方的第一个表格 block_id"""
    blocks = api_get(token, f"/docx/v1/documents/{doc_id}/blocks?page_size=500")
    items = blocks.get("data", {}).get("items", [])

    heading_index = -1
    for i, b in enumerate(items):
        text = ""
        for e in b.get("text", {}).get("elements", []):
            text += e.get("text_run", {}).get("content", "")
        if text == heading_text:
            heading_index = i
            break

    if heading_index == -1:
        return None

    for b in items[heading_index + 1:]:
        if b.get("block_type") == 31:
            return b["block_id"]

    return None


def get_cell_block_ids(token, doc_id, table_block_id):
    """
    获取表格中所有单元格的 block_id，返回二维数组 [row][col]
    飞书表格结构：table > row > cell，每个 row 只有一个 cell
    需要按列数分组。
    """
    rows = api_get_all_children(token, doc_id, table_block_id)

    all_cells = []
    for row in rows:
        try:
            cell_items = api_get_all_children(token, doc_id, row["block_id"])
        except Exception:
            continue
        if cell_items:
            all_cells.append(cell_items[0]["block_id"])

    return all_cells


def fill_cells(token, doc_id, cell_ids, col_count, knowledge_items, tl_rows, comp_rows):
    """回填统计结果到表格"""
    ok = 0
    fail = 0

    for item_idx, item in enumerate(knowledge_items):
        word = item.get("word", "")
        pattern = item.get("pattern", "")

        # 计算该行在 cell_ids 中的起始位置
        # 跳过表头行
        row_start = (item_idx + 1) * col_count

        if word:
            word_only = word.split()[0] if word else ""
            w_in = count_occurrences(word_only, tl_rows)
            w_out = count_occurrences(word_only, comp_rows)

            # col 1 = 输入(单词)
            if row_start + 1 < len(cell_ids):
                ok += patch_cell(token, doc_id, cell_ids[row_start + 1], str(w_in))
                time.sleep(0.3)
            else:
                fail += 1
            # col 2 = 输出(单词)
            if row_start + 2 < len(cell_ids):
                ok += patch_cell(token, doc_id, cell_ids[row_start + 2], str(w_out))
                time.sleep(0.3)
            else:
                fail += 1

        if pattern:
            p_in = count_pattern_occurrences(pattern, tl_rows)
            p_out = count_pattern_occurrences(pattern, comp_rows)

            # col 4 = 输入(句型)
            if row_start + 4 < len(cell_ids):
                ok += patch_cell(token, doc_id, cell_ids[row_start + 4], str(p_in))
                time.sleep(0.3)
            else:
                fail += 1
            # col 5 = 输出(句型)
            if row_start + 5 < len(cell_ids):
                ok += patch_cell(token, doc_id, cell_ids[row_start + 5], str(p_out))
                time.sleep(0.3)
            else:
                fail += 1

    print(f"  回填完成: {ok} 成功, {fail} 失败")
    return ok, fail


def patch_cell(token, doc_id, block_id, content):
    """更新单个文本块的内容"""
    body = {
        "update_text_elements": {
            "elements": [{"text_run": {"content": content}}]
        }
    }
    try:
        resp = api_patch(token, f"/docx/v1/documents/{doc_id}/blocks/{block_id}", body)
        return resp.get("code") == 0
    except Exception as e:
        print(f"  ❌ 回填失败 {block_id}: {e}")
        return False


# ── 报告 ──────────────────────────────────────────────

def print_report(knowledge_items, tl_rows, comp_rows):
    """打印统计报告"""
    print("\n" + "=" * 60)
    print("📊 知识点覆盖率统计")
    print("=" * 60)

    print(f"\nTL 行（输入）: {len(tl_rows)} 行")
    print(f"组件行（输出）: {len(comp_rows)} 行")

    print(f"\n{'知识点':<30} {'输入':>4} {'输出':>4}")
    print("-" * 42)

    for item in knowledge_items:
        word = item.get("word", "")
        pattern = item.get("pattern", "")

        if word:
            word_only = word.split()[0]
            w_in = count_occurrences(word_only, tl_rows)
            w_out = count_occurrences(word_only, comp_rows)
            print(f"{word:<30} {w_in:>4} {w_out:>4}")

        if pattern:
            p_in = count_pattern_occurrences(pattern, tl_rows)
            p_out = count_pattern_occurrences(pattern, comp_rows)
            label = pattern[:28] + ".." if len(pattern) > 30 else pattern
            print(f"  {label:<28} {p_in:>4} {p_out:>4}")

    # 质量检查
    print("\n" + "=" * 60)
    print("🔍 用法质量快速检查")
    print("=" * 60)

    for item in knowledge_items:
        word = item.get("word", "")
        if not word:
            continue
        word_only = word.split()[0]
        w_out = count_occurrences(word_only, comp_rows)
        w_total = count_occurrences(word_only, tl_rows) + w_out

        if w_total < 2:
            print(f"\n⚠️  {word} — 仅出现 {w_total} 次，建议增加曝光")
        elif w_total > 8:
            print(f"\n💡 {word} — 出现 {w_total} 次，可适当精简")


# ── 主入口 ────────────────────────────────────────────

def main():
    if len(sys.argv) < 2:
        print("用法: python3 knowledge_coverage_analyzer.py <飞书文档ID或URL>")
        sys.exit(1)

    doc_id = parse_doc_id(sys.argv[1])
    print(f"📄 文档 ID: {doc_id}")

    # Step 1: 用 lark-cli 获取 markdown
    print("📥 获取文档内容...")
    md = fetch_markdown(doc_id)

    # Step 2: 解析知识点现状表格
    print("📋 解析「知识点现状」表格...")
    knowledge_data = parse_markdown_table(md, "知识点现状")
    if not knowledge_data:
        print("❌ 未找到「知识点现状」表格")
        sys.exit(1)

    # 表头: 单词, 输入, 输出, 句型, 输入, 输出
    knowledge_items = []
    for row in knowledge_data[1:]:  # 跳过表头
        if len(row) < 6:
            continue
        word = row[0]
        # 清理 markdown 格式：去掉列表符号
        word = re.sub(r'^[-*+]\s+', '', word).strip()
        pattern = row[3].strip()
        if not word and not pattern:
            continue
        knowledge_items.append({"word": word, "pattern": pattern})

    print(f"  知识点: {len(knowledge_items)} 条")
    for item in knowledge_items:
        w = item["word"] or "(无)"
        p = item["pattern"] or "(无)"
        print(f"    - {w} | {p}")

    # Step 3: 解析剧本主表
    print("\n📋 解析剧本主表...")
    script_data = find_table_by_first_cell_md(md, "类型")
    if not script_data:
        print("❌ 未找到剧本主表（第一列第一行为「类型」）")
        sys.exit(1)

    tl_rows = []
    comp_rows = []
    last_was_tl = True
    for row in script_data[1:]:  # 跳过表头
        if not row:
            continue
        type_text = row[0].strip()
        # 排除「知识点」列（第6列，index 5）避免单词自计数
        row_text = " ".join(row[1:5] + row[6:] if len(row) > 6 else row[1:5])
        # 分类：类型列明确标记为 TL → 输入；
        # 类型列包含交互关键词 → 输出；
        # 空类型 → 继承上一行
        if type_text.startswith("TL"):
            tl_rows.append(row_text)
            last_was_tl = True
        elif not type_text:
            if last_was_tl:
                tl_rows.append(row_text)
            else:
                comp_rows.append(row_text)
        elif any(kw in type_text for kw in ("对话朗读", "对话挖空", "剧情任务", "对话", "挖空", "互动")):
            comp_rows.append(row_text)
            last_was_tl = False
        else:
            # 其他（场景描述等）归为 TL
            tl_rows.append(row_text)
            last_was_tl = True

    print(f"  TL 行: {len(tl_rows)}, 组件行: {len(comp_rows)}")

    # Step 4: 获取 token + 回填
    print("\n🔑 获取 API Token...")
    token = get_token()

    print("📝 查找表格 block ID...")
    knowledge_table_id = find_table_block_id(token, doc_id, "知识点现状")
    if not knowledge_table_id:
        print("❌ 未找到「知识点现状」表格 block")
        sys.exit(1)

    cell_ids = get_cell_block_ids(token, doc_id, knowledge_table_id)
    col_count = 6  # 知识点现状表格固定6列
    print(f"  表格 block: {knowledge_table_id}, 共 {len(cell_ids)} 个单元格")

    print("\n📝 回填统计结果...")
    fill_cells(token, doc_id, cell_ids, col_count, knowledge_items, tl_rows, comp_rows)

    # Step 5: 打印报告
    print_report(knowledge_items, tl_rows, comp_rows)

    print("\n✅ 完成！")


if __name__ == "__main__":
    main()