#!/usr/bin/env python3
"""
小唯消息面情感评分 — 东方财富公告 + 关键词情感词典
==================================================
抓取关注股票最新公告标题，用情感词典打分。

评分规则：
  利空词（减持/亏损/违规/立案/质押/诉讼/退市/警示/下跌/处罚）→ -1
  利好词（增持/回购/分红/中标/增长/突破/合作/扩产/盈利/上调/签约）→ +1
  中性 → 0
  综合多个公告后 clip 到 [-1, 1]

用法：
  python3 stock_sentiment.py            # 输出情感评分
  python3 stock_sentiment.py --json     # JSON 输出（供矛盾周报引用）
"""
import json, sys, subprocess, shlex, os, re
from datetime import datetime
from pathlib import Path

OUTPUT = Path.home() / ".hermes" / "stock_backtest"
OUTPUT.mkdir(exist_ok=True)

# 关注股票（代码, 名称, 行业）
WATCHED = [
    ("000858", "五粮液", "白酒"),
    ("600519", "贵州茅台", "白酒"),
    ("000568", "泸州老窖", "白酒"),
    ("002304", "洋河股份", "白酒"),
    ("600036", "招商银行", "银行"),
    ("601318", "中国平安", "保险"),
    ("000001", "平安银行", "银行"),
    ("300750", "宁德时代", "新能源"),
    ("002594", "比亚迪", "新能源"),
    ("002415", "海康威视", "科技"),
    ("601088", "中国神华", "煤炭"),
    ("688981", "中芯国际", "半导体"),
    ("600030", "中信证券", "证券"),
    ("000333", "美的集团", "家电"),
]

# 情感词典
NEGATIVE_WORDS = [
    "减持", "亏损", "违规", "立案", "质押", "诉讼", "退市", "警示",
    "下跌", "处罚", "风险", "终止", "暂停", "下滑", "恶化", "逾期",
    "冻结", "查封", "调查", "降级", "下调", "失败", "延期", "变卖",
]
POSITIVE_WORDS = [
    "增持", "回购", "分红", "中标", "增长", "突破", "合作", "扩产",
    "盈利", "上调", "签约", "创新高", "预增", "扭亏", "获批", "落地",
    "推出", "发布", "投资", "签订", "完成", "超预期", "翻倍", "新签订单",
]

# 中性词（公告常见但无方向性）
NEUTRAL_WORDS = ["会议", "报告", "公告", "章程", "制度", "通知", "更正", "说明"]


def get_url(url, timeout=8):
    env = dict(os.environ)
    for k in ["http_proxy", "https_proxy", "HTTP_PROXY", "HTTPS_PROXY"]:
        env.pop(k, None)
    cmd = f"curl -s --max-time {timeout} --compressed {shlex.quote(url)}"
    try:
        r = subprocess.run(cmd, shell=True, capture_output=True, timeout=timeout + 2, env=env)
        return r.stdout.decode("utf-8", errors="ignore")
    except Exception:
        return ""


def get_announcements(code, limit=5):
    """东方财富个股公告（标题列表）"""
    url = (f"https://np-anotice-stock.eastmoney.com/api/security/ann"
           f"?sr=-1&page_size={limit}&page_index=1&ann_type=A"
           f"&client_source=web&stock_list={code}&f_node=0&s_node=0")
    text = get_url(url)
    if not text or len(text) < 50:
        return []
    try:
        data = json.loads(text)
        items = data.get("data", {}).get("list", [])
        titles = []
        for item in items:
            title = item.get("title", "").strip()
            # 清理 HTML 标签
            title = re.sub(r"<[^>]+>", "", title)
            if title:
                titles.append(title)
        return titles
    except Exception:
        return []


def score_title(title):
    """单条标题情感评分：返回 (score, matched)"""
    pos_hits = [w for w in POSITIVE_WORDS if w in title]
    neg_hits = [w for w in NEGATIVE_WORDS if w in title]
    if pos_hits and not neg_hits:
        return 1, pos_hits
    if neg_hits and not pos_hits:
        return -1, neg_hits
    if pos_hits and neg_hits:
        return 0, pos_hits + neg_hits  # 混合中性
    return 0, []


def scan_sentiment():
    """扫描所有关注股票，返回情感评分"""
    results = []
    print("=" * 60)
    print(f"小唯消息面情感扫描 {datetime.now().strftime('%Y-%m-%d %H:%M')}")
    print("=" * 60)

    for code, name, industry in WATCHED:
        titles = get_announcements(code, limit=5)
        if not titles:
            print(f"  ⚠️ {name}: 无公告数据")
            results.append({"code": code, "name": name, "industry": industry,
                            "score": 0, "titles": [], "reasons": ["无公告"]})
            continue

        scores = [score_title(t) for t in titles]
        # 有实质方向的最新公告优先（取最近3条的平均）
        recent = [s for s, _ in scores[:3]]
        avg = sum(recent) / len(recent) if recent else 0
        final = max(-1, min(1, avg))

        hits = []
        for t, (s, words) in zip(titles[:3], scores[:3]):
            if words:
                emoji = "🔴" if s < 0 else ("🟢" if s > 0 else "⚪")
                hits.append(f"{emoji}{t[:35]}")

        results.append({"code": code, "name": name, "industry": industry,
                        "score": final, "titles": titles[:3], "hits": hits,
                        "reasons": hits})
        emoji = "🔴" if final < 0 else ("🟢" if final > 0 else "⚪")
        print(f"  {emoji} {name}({code}) [{industry}] 情感{final:+.0f}")

    # 汇总：全市场情感（消息面评分）
    valid = [r for r in results if r.get("titles")]
    if valid:
        avg_all = sum(r["score"] for r in valid) / len(valid)
        message = max(-1, min(1, avg_all))
    else:
        message = 0

    print(f"\n📊 市场消息面综合评分: {message:+.0f}")

    out = OUTPUT / "sentiment_scan.json"
    with open(out, "w", encoding="utf-8") as f:
        json.dump({"generated": datetime.now().strftime("%Y-%m-%d %H:%M"),
                   "message_score": message,
                   "stocks": results}, f, ensure_ascii=False, indent=2)
    print(f"📁 已保存: {out}")

    if "--json" in sys.argv:
        return message
    return results


if __name__ == "__main__":
    scan_sentiment()