import argparse
import csv
import os
from collections import Counter, defaultdict
from datetime import datetime, timezone, timedelta
from pathlib import Path


def read_csv(path):
    with path.open("r", encoding="utf-8-sig", newline="") as handle:
        return list(csv.DictReader(handle))


def top(counter, n=10):
    return counter.most_common(n)


def safe_cell(value):
    return (value or "").replace("\n", " ").strip()


def relative_link(target, base_dir):
    return Path(os.path.relpath(target, base_dir)).as_posix()


COMPANY_STOPWORDS = {
    "有色金属",
    "能源金属",
    "工业金属",
    "基本金属",
    "贵金属",
    "小金属",
    "金属新材料",
    "中重稀土",
    "矿产资源",
    "自然资源",
    "稀土资源",
    "COMEX黄金",
    "LME铜",
    "LME铝",
    "国家能源",
    "中国新能源",
    "基础化工",
    "稀土冶炼",
    "年全球稀土",
    "全球稀土",
}


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--project-root", default=".")
    parser.add_argument("--outputs-dir", required=True)
    parser.add_argument("--result-index", required=True)
    parser.add_argument("--fact-files", nargs="+", required=True)
    parser.add_argument("--company-files", nargs="+", required=True)
    parser.add_argument("--snapshot", nargs="+", required=True)
    args = parser.parse_args()

    project_root = Path(args.project_root).resolve()
    outputs_dir = project_root / args.outputs_dir
    outputs_dir.mkdir(parents=True, exist_ok=True)
    result_index = project_root / args.result_index
    created_at = datetime.now(timezone(timedelta(hours=8))).isoformat(timespec="seconds")

    facts = []
    for file_name in args.fact_files:
        path = project_root / file_name
        if path.exists():
            facts.extend(read_csv(path))

    companies = []
    for file_name in args.company_files:
        path = project_root / file_name
        if path.exists():
            companies.extend(read_csv(path))

    snapshot = []
    for snapshot_file in args.snapshot:
        path = project_root / snapshot_file
        if path.exists():
            snapshot.extend(read_csv(path))

    metal_counter = Counter()
    theme_counter = Counter()
    metal_theme = defaultdict(Counter)
    example_by_metal = defaultdict(list)
    for row in facts:
        metals = [tag for tag in row.get("metal_tags", "").split("|") if tag]
        themes = [tag for tag in row.get("theme_tags", "").split("|") if tag]
        for metal in metals:
            metal_counter[metal] += 1
            if len(example_by_metal[metal]) < 4:
                example_by_metal[metal].append(row)
            for theme in themes:
                metal_theme[metal][theme] += 1
        for theme in themes:
            theme_counter[theme] += 1

    company_counter = Counter()
    company_examples = {}
    for row in companies:
        name = row.get("company_name", "")
        if (
            not name
            or len(name) > 14
            or name in COMPANY_STOPWORDS
            or name.isdigit()
            or "本报告" in name
            or "所有材料" in name
        ):
            continue
        company_counter[name] += 1
        company_examples.setdefault(name, row)

    by_metal_snapshot = defaultdict(list)
    for row in snapshot:
        by_metal_snapshot[row.get("metal", "")].append(row)

    industry_lines = [
        "# 有色行业视图草稿",
        "",
        "状态：DRAFT_FOR_REVIEW",
        f"生成时间：{created_at}",
        "",
        "## 证据边界",
        "",
        "本文档由 BATCH-001 与 BATCH-003 的规则抽取事实句、BATCH-003 转换/抽取 manifest、外部价格库存快照共同生成。当前尚未完成关键事实段落、表格定位和人工复核，因此只作为核心文档草稿，不输出正式行业结论或交易建议。",
        "",
        "## 材料覆盖",
        "",
        f"- 规则抽取事实句：{len(facts)} 条",
        f"- 公司提及草表：{len(companies)} 条",
        "- BATCH-003 转换状态：329 个补登记 raw 中 309 个已转换、20 个进入转换缺口。",
        "",
        "## 金属主题分布",
        "",
        "| 金属 | 事实句数量 | 主要主题 |",
        "|---|---:|---|",
    ]
    for metal, count in top(metal_counter, 12):
        themes = ", ".join([f"{name}:{cnt}" for name, cnt in top(metal_theme[metal], 5)])
        industry_lines.append(f"| {metal} | {count} | {themes} |")
    industry_lines.extend(["", "## 初步行业读法", ""])
    for metal, _ in top(metal_counter, 8):
        themes = [name for name, _ in top(metal_theme[metal], 4)]
        industry_lines.append(f"### {metal}")
        industry_lines.append("")
        industry_lines.append(f"- 抽取材料中，{metal}相关句子主要集中在：{', '.join(themes) if themes else '待复核'}。")
        if example_by_metal[metal]:
            industry_lines.append("- 待复核样例：")
            for row in example_by_metal[metal][:2]:
                industry_lines.append(f"  - {safe_cell(row.get('evidence_text'))}（{row.get('doc_id')}，{row.get('source_location')}）")
        industry_lines.append("")
    industry_lines.extend(
        [
            "## 下一步复核点",
            "",
            "1. 对价格、库存、供给、项目投产和政策相关事实补段落或表格定位。",
            "2. 将低置信度公司提及草表收敛为公司证据卡，补业务占比、资源/产能、成本位置和触发/失效条件。",
            "3. 外部价格库存快照需要按来源继续补 SHFE、COMEX、SMM 等可复核数值。",
        ]
    )

    market_lines = [
        "# 有色市场视图草稿",
        "",
        "状态：DRAFT_FOR_REVIEW",
        f"生成时间：{created_at}",
        "",
        "## 证据边界",
        "",
        "本文档仅汇总公开外部快照和研报抽取中的市场相关句子。涉及实时行情、新闻、市场表现或外部信息的内容按 supplement/evidence 分账，不作为正式交易判断。",
        "",
        "## 外部价格库存快照",
        "",
        "| 金属 | 指标 | 数值 | 单位 | 市场/地区 | 来源 |",
        "|---|---|---:|---|---|---|",
    ]
    for row in snapshot:
        market_lines.append(
            f"| {row.get('metal')} | {row.get('indicator')} | {row.get('value')} | {row.get('unit')} | {row.get('market_or_region')} | {row.get('source_name')} |"
        )
    market_lines.extend(["", "## 市场变量分布", "", "| 主题 | 事实句数量 |", "|---|---:|"])
    for theme, count in top(theme_counter, 12):
        market_lines.append(f"| {theme} | {count} |")
    market_lines.extend(
        [
            "",
            "## 观察框架",
            "",
            "- 工业金属：优先跟踪价格、库存、冶炼/矿端供应、下游开工和宏观需求预期之间是否一致。",
            "- 贵金属：区分金银价格的宏观利率/美元逻辑与材料端白银需求逻辑。",
            "- 能源金属：锂、镍、钴需要把价格方向与库存、排产、产能释放和政策扰动分开验证。",
            "- 稀土/小金属：重点核对供给管制、出口、库存和高端制造需求，不用单一价格点推导结论。",
            "",
            "## 缺口",
            "",
            "1. SHFE 库存与国内现货价格仍需补可复核数值入口。",
            "2. 锂和稀土的部分价格为补充口径，仍需 SMM 或交易所/行业协会直接数值复核。",
            "3. 市场异常表现、新闻事件和政策事件需要后续按 darkline 分流形成独立证据链。",
        ]
    )

    company_lines = [
        "# 有色公司视图草稿",
        "",
        "状态：DRAFT_FOR_REVIEW",
        f"生成时间：{created_at}",
        "",
        "## 证据边界",
        "",
        "公司视图当前来自规则公司名扫描，置信度较低，只用于确定后续证据卡优先级。不得据此输出交易指令、收益承诺或无证据强结论。",
        "",
        "## 公司提及频次",
        "",
        "| 公司名片段 | 提及次数 | 样例定位 | 样例上下文 |",
        "|---|---:|---|---|",
    ]
    for name, count in top(company_counter, 40):
        row = company_examples[name]
        company_lines.append(
            f"| {name} | {count} | {row.get('doc_id')} / {row.get('source_location')} | {safe_cell(row.get('context_snippet'))[:120]} |"
        )
    company_lines.extend(
        [
            "",
            "## 后续公司证据卡字段",
            "",
            "1. 公司简称、证券代码、主营金属、核心资产或业务环节。",
            "2. 资源量、产能、成本位置、扩产/投产节奏、价格敏感性。",
            "3. 观察条件、触发条件、失效条件和主要风险。",
            "4. 原文页码/段落/表格定位、外部公告或数据来源、review_status。",
        ]
    )

    industry_path = outputs_dir / "nonferrous_industry_view_draft.md"
    market_path = outputs_dir / "nonferrous_market_view_draft.md"
    company_path = outputs_dir / "nonferrous_company_view_draft.md"
    industry_path.write_text("\n".join(industry_lines) + "\n", encoding="utf-8")
    market_path.write_text("\n".join(market_lines) + "\n", encoding="utf-8")
    company_path.write_text("\n".join(company_lines) + "\n", encoding="utf-8")

    result_lines = [
        "# ANA-YS-INDUSTRY-001 结果入口",
        "",
        "状态：DRAFT_FOR_REVIEW",
        f"更新时间：{created_at}",
        "",
        "## 核心文档",
        "",
        f"1. [有色行业视图草稿]({relative_link(industry_path, result_index.parent)})",
        f"2. [有色市场视图草稿]({relative_link(market_path, result_index.parent)})",
        f"3. [有色公司视图草稿]({relative_link(company_path, result_index.parent)})",
        "",
        "## 关键过程产物",
        "",
        "- `ana-data/cases/有色案例/ANA-YS-INDUSTRY-001/outputs/batch003_conversion_summary.md`",
        "- `ana-data/cases/有色案例/ANA-YS-INDUSTRY-001/outputs/batch003_extract001_summary.md`",
        "- `ana-data/cases/有色案例/ANA-YS-INDUSTRY-001/outputs/key_fact_review_batch003_pass001_summary.md`",
        "- `ana-data/cases/有色案例/ANA-YS-INDUSTRY-001/outputs/company_evidence_card_batch003_pass001_summary.md`",
        "- `ana-data/cases/有色案例/ANA-YS-INDUSTRY-001/outputs/price_inventory_supplement_pass002_summary.md`",
        "- `ana-data/cases/有色案例/supplement/nonferrous_price_inventory_snapshot_20260625.csv`",
        "- `ana-data/cases/有色案例/supplement/nonferrous_price_inventory_supplement_pass002_20260625.csv`",
        "- `ana-data/cases/有色案例/manifest/conversion_gap_batch003.csv`",
        "",
        "## 限制",
        "",
        "BATCH-001 与 BATCH-003 关键事实尚未完成段落/表格定位和人工复核。对外正式结论、公司投资读法或交付前，必须提交执行审核或输出审核。",
    ]
    result_index.parent.mkdir(parents=True, exist_ok=True)
    result_index.write_text("\n".join(result_lines) + "\n", encoding="utf-8")

    print(f"industry={industry_path.relative_to(project_root).as_posix()}")
    print(f"market={market_path.relative_to(project_root).as_posix()}")
    print(f"company={company_path.relative_to(project_root).as_posix()}")
    print(f"result_index={result_index.relative_to(project_root).as_posix()}")


if __name__ == "__main__":
    main()