tradingSystem/scripts/probe_plan_api.py

# -*- coding: utf-8 -*-
"""
上游选股计划接口实机探活 (只读, 不写任何表)
============================================
    docker compose run --rm --no-deps pms-web python scripts/probe_plan_api.py
    docker compose run --rm --no-deps pms-web python scripts/probe_plan_api.py --date 2026-07-29
    docker compose run --rm --no-deps pms-web python scripts/probe_plan_api.py --top 30 --with-price

干什么: 拿真实应答验证五件事 —— ① 接口通不通、② 字段口径与单测 fixture 是否一致、
③ 按当前参数筛出来的候选池长什么样、④ (--with-price) 这些票行情里到底有没有价、
⑤ 跟库里最新那份榜比, 谁新进谁掉榜、持仓票有没有被摘。
第 ④ 项最容易翻车: 计划不带价格, 价格取不到的票会在候选池里被静默剔除。

只读: 不会 upsert pms_industry_map (那由 /api/ops/plan-refresh 或盘前调度做), 也不下单。
第 ⑤ 项默认也是只读 (拿手上这份跟库里最新快照比, 不写库); **唯一的写操作**是显式加
`--snapshot` 时落一行 pms_plan_snapshot。
"""
from __future__ import annotations

import argparse
import json
import os
import sys

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from app.services import plan_feed as pf                                       # noqa: E402


def _fmt(v, n=4):
    return "-" if v is None else (f"{v:.{n}f}" if isinstance(v, float) else str(v))


def _pct(v):
    """upside 是相对现价的比例: 2.1203 → +212% (与上游 format=md 的「预期空间」同口径)。"""
    return "-" if v is None else f"{v * 100:+.0f}%"


def _default_gateway() -> str:
    """容器默认网关 = 宿主机在网桥上的地址。不依赖 ip/route 命令, 直接读 /proc。"""
    try:
        with open("/proc/net/route", encoding="utf-8") as f:
            for line in f.readlines()[1:]:
                p = line.split()
                if len(p) > 2 and p[1] == "00000000" and p[2] != "00000000":
                    h = p[2]
                    return ".".join(str(int(h[i:i + 2], 16)) for i in (6, 4, 2, 0))
    except Exception:
        pass
    return ""


def _try_bases(ps, args) -> int:
    """逐个试候选地址。容器 → 宿主机这条路最容易断, 这里一次把可能的走法全试掉。

    ConnectTimeout  = 包被丢了 (宿主机 INPUT 链拦了网桥进来的流量), 不是没人监听
    ConnectionError = 端口没人听 (地址对了但服务不在这台/没起)
    """
    port = "8300"
    cfg = (ps.get("PMS_PLAN_API_BASE", "") or "").strip().rstrip("/")
    gw = _default_gateway()
    cands = []
    for b in ([args.base] if args.base else []) + [
            cfg, f"http://host.docker.internal:{port}",
            (f"http://{gw}:{port}" if gw else ""), f"http://172.17.0.1:{port}",
            f"http://192.168.16.155:{port}", f"http://127.0.0.1:{port}"]:
        b = (b or "").strip().rstrip("/")
        if b and b not in cands:
            cands.append(b)
    print(f"[探路] 容器默认网关 = {gw or '(读不到)'}; 当前配置 = {cfg or '(空)'}")
    print(f"       共 {len(cands)} 个候选, 每个 3 秒超时")
    ok = []
    for b in cands:
        tag = "  当前配置" if b == cfg else ""
        try:
            plan = pf.fetch(base=b, path=ps.get("PMS_PLAN_API_PATH", "/plan"), timeout=3)
            print(f"  通    {b}{tag}   → 计划 {plan['date']}, 主榜 {plan['returned']['main']} 条")
            ok.append(b)
        except pf.PlanFeedError as e:
            msg = str(e).split(": ", 1)[-1]
            print(f"  不通  {b}{tag}   → {msg[:150]}")
    print("-" * 62)
    if not ok:
        print("全都不通。按下面顺序排:")
        print("  1. 宿主机上确认在听:  ss -ltnp | grep 8300")
        print("  2. 若上面显示 127.0.0.1:8300 —— 服务只绑了 loopback, 容器永远进不来,")
        print("     要么让它绑 0.0.0.0, 要么把 PMS 容器改 network_mode: host (不推荐)。")
        print("  3. 若显示 0.0.0.0:8300 而这里仍超时 —— 宿主机 INPUT 链拦了网桥流量, 放行:")
        print("       sudo firewall-cmd --permanent --zone=trusted --add-source=172.16.0.0/12")
        print("       sudo firewall-cmd --reload")
        print("     (或 iptables: sudo iptables -I INPUT -s 172.16.0.0/12 -p tcp --dport 8300 -j ACCEPT)")
        print("  4. 若计划服务本身也在 docker 里 —— 最干净的是把它和 pms-* 放同一个 network,")
        print("     base 直接填 http://<那个容器名>:8300, 不绕宿主机。")
        return 1
    print(f"能通的地址: {ok[0]}")
    print(f"→ 页面「参数设置」把 PMS_PLAN_API_BASE 改成它, 或先跑:")
    print(f"   curl -s -X POST http://127.0.0.1:38100/api/params -H 'Content-Type: application/json' \\")
    print(f"     -d '{{\"key\":\"PMS_PLAN_API_BASE\",\"value\":\"{ok[0]}\"}}'")
    return 0


LIMIT_KEYS = ["limit", "top", "topn", "top_n", "n", "size", "page_size", "per_page",
              "count", "max", "rows", "num", "full"]


def _try_limit(ps, args) -> int:
    """探"取全量"的参数名。

    上游默认只回主榜 20 条, 而 counts.main=961 —— 候选池实际只在这 20 条里选, 且那 20 条
    已经被上游按自己的「每主题限额 5」裁过。参数名探出来填进 PMS_PLAN_QUERY_EXTRA 即可,
    不用改代码。全都不认就只能等上游给分页方式 (UPSTREAM_PLAN_API.md Q3)。
    """
    base = (args.base or ps.get("PMS_PLAN_API_BASE", "") or "").strip().rstrip("/")
    path = ps.get("PMS_PLAN_API_PATH", "/plan")
    if not base:
        print("PMS_PLAN_API_BASE 为空, 先把地址配上 (或用 --base)")
        return 2
    try:
        base_plan = pf.fetch(base=base, path=path, timeout=15, date=args.date)
    except pf.PlanFeedError as e:
        print(f"基准请求就失败了: {e}")
        return 1
    n0, c0 = base_plan["returned"]["main"], base_plan["counts"]["main"]
    print(f"[基准] 不带任何参数: 主榜回 {n0} 条 (上游 counts={c0})")
    if not base_plan["truncated"]["main"]:
        print("      没有截断 —— 已经是全量, 不需要分页参数。")
        return 0
    print(f"[探参] 依次试 {len(LIMIT_KEYS)} 个常见参数名, 取值 {args.limit_value}")
    hits = []
    for k in LIMIT_KEYS:
        try:
            p = pf.fetch(base=base, path=path, timeout=20, date=args.date,
                         extra_params={k: args.limit_value})
            n = p["returned"]["main"]
            flag = "  ← 生效" if n > n0 else ""
            print(f"  {k:<10}= {args.limit_value:<6} → 主榜 {n} 条{flag}")
            if n > n0:
                hits.append((k, n))
        except pf.PlanFeedError as e:
            print(f"  {k:<10}= {args.limit_value:<6} → 报错: {str(e).split(': ', 1)[-1][:90]}")
    print("-" * 62)
    if not hits:
        print(f"没有一个参数名生效, 主榜始终 {n0} 条。")
        print("→ 这条只能问上游: 取全量 961 条的正确方式是什么 (分页? 另一个端点?)")
        print("   见 UPSTREAM_PLAN_API.md Q3。在此之前候选池就是从这 20 条里选,")
        print("   并且要知道: 这 20 条已被上游按「每主题限额 5」裁过, PMS 侧的行业约束")
        print("   是在一个已经被裁过的池子上再裁一次。")
        return 1
    hits.sort(key=lambda x: -x[1])
    k, n = hits[0]
    print(f"生效: {k}={args.limit_value} → 主榜 {n} 条 (原 {n0})")
    print("→ 页面「参数设置」把 PMS_PLAN_QUERY_EXTRA 填成下面这串, 或直接:")
    print(f"   curl -s -X POST http://127.0.0.1:38100/api/params -H 'Content-Type: application/json' \\")
    print(f"     -d '{{\"key\":\"PMS_PLAN_QUERY_EXTRA\",\"value\":\"{k}={args.limit_value}\"}}'")
    return 0


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--date", default=None, help="指定计划日期 YYYY-MM-DD (缺省=上游最新)")
    ap.add_argument("--top", type=int, default=20, help="打印主榜前 N 行")
    ap.add_argument("--observe", action="store_true", help="连观察档一起打印")
    ap.add_argument("--with-price", action="store_true",
                    help="逐只查行情价 (需 Redis 通; 用来看候选池会不会因无价掉票)")
    ap.add_argument("--base", default=None,
                    help="临时覆盖 PMS_PLAN_API_BASE (只影响本次, 不落库)")
    ap.add_argument("--try-bases", action="store_true",
                    help="逐个试候选地址, 报告哪个能通 —— 容器连不上宿主机时先跑这个")
    ap.add_argument("--try-limit", action="store_true",
                    help="探取全量的参数名 (上游默认只回主榜 20 条, counts 却是 961)")
    ap.add_argument("--limit-value", default="1000", help="--try-limit 用的取值 (默认 1000)")
    ap.add_argument("--snapshot", action="store_true",
                    help="把这份名册落进 pms_plan_snapshot (**本脚本唯一的写操作**, 默认不写)")
    ap.add_argument("--json", action="store_true", help="原样打印解析后的结构")
    args = ap.parse_args()

    from app.services import param_store as ps
    if args.try_bases:
        return _try_bases(ps, args)
    if args.try_limit:
        return _try_limit(ps, args)
    base = (args.base or ps.get("PMS_PLAN_API_BASE", "") or "").strip()
    print(f"[1] 接口配置  base={base or '(空 —— 候选池将恒为空)'} "
          f"path={ps.get('PMS_PLAN_API_PATH', '/plan')} "
          f"timeout={ps.get_int('PMS_PLAN_TIMEOUT', 10)}s")
    if not base:
        print("    → PMS_PLAN_API_BASE 为空。页面「参数设置」填上再跑。")
        return 2

    q = pf.effective_query()
    print(f"    发出的查询参数 {q or '(无 —— 将吃上游默认 top=20/obs_top=10/theme_cap=5)'}")
    try:
        plan = pf.fetch(date=args.date, base=base, extra_params=q)
    except pf.PlanFeedError as e:
        print(f"[2] 取数失败: {e}")
        return 1
    print(f"[2] 取数成功  {plan['url']}")
    print(f"    date={plan['date']}  heat_date={plan['heat_date']}  "
          f"snapshot={plan['market_snapshot_days']}  theme_cap={plan['theme_cap']}")
    print(f"    请求参数 {plan['requested']}")
    f = plan["funnel"]
    print(f"    漏斗: 打分池 主榜 {f['scored_main']} / 观察 {f['scored_observe']}"
          f"  →(券商预期 + 目标价≥现价 + 主题限额 + top)→"
          f"  实收 主榜 {f['returned_main']} / 观察 {f['returned_observe']}")
    if plan["truncated"]["main"]:
        print(f"    · 主榜正好吃满 top={plan['requested']['top']} —— 要不要紧看 [6] 的"
              f"「白名单档位取全」那一行")
    else:
        print(f"    主榜没吃满 top={plan['requested']['top']}, 说明这就是上游能给的全部"
              f" (受主题限额 {plan['theme_cap']} 与价格筛限制)")

    age = pf.plan_age_tdays(plan["date"])
    limit = ps.get_int("PMS_PLAN_STALE_TDAYS", 1)
    verdict = "新鲜" if age <= limit else f"**过期** (上限 {limit}, 候选池会拒用)"
    print(f"[3] 新鲜度    日龄 {age} 个交易日 → {verdict}")

    print(f"[4] 主榜前 {args.top} (score 降序即上游 rank 序)")
    hdr = f"    {'rank':>4} {'代码':<11} {'名称':<8} {'score':>8} {'档位':<6} {'主题':<10} " \
          f"{'heat':>7} {'预期空间':>9}"
    print(hdr + ("   " + "行情价" if args.with_price else ""))
    rows = plan["main"][:max(1, args.top)]
    if args.observe:
        rows += plan["observe"][:max(1, args.top)]
    prices = {}
    if args.with_price:
        from app.services import market
        for r in rows:
            try:
                pp = market.plan_price(r["ts_code"])
                prices[r["ts_code"]] = ((f"{pp['price']}" if pp["source"] == "realtime"
                                         else f"{pp['price']} (昨收)") if pp["price"]
                                        else "**无价(会被剔除)**")
            except Exception as e:
                prices[r["ts_code"]] = f"ERR {type(e).__name__}"
    for r in rows:
        line = (f"    {str(r['rank']):>4} {r['ts_code']:<11} {(r['name'] or '-'):<8} "
                f"{_fmt(r['score'], 2):>8} {(r['tier'] or '观察'):<6} "
                f"{(r['theme'] or '-'):<10} {_fmt(r['heat']):>7} {_pct(r['upside']):>9}")
        if args.with_price:
            line += f"   {prices.get(r['ts_code']) or '-'}"
        print(line)

    themes = {}
    for r in plan["main"]:
        themes[r["theme"] or "(无)"] = themes.get(r["theme"] or "(无)", 0) + 1
    top_themes = sorted(themes.items(), key=lambda x: -x[1])[:12]
    print(f"[5] 主榜主题分布 (共 {len(themes)} 个主题): " +
          ", ".join(f"{k}×{v}" for k, v in top_themes))

    sel = pf.select_candidates(
        plan, top_n=ps.get_int("PMS_PLAN_TOP_N", 30), tiers=ps.get_list("PMS_PLAN_TIERS", []),
        include_observe=ps.get_bool("PMS_PLAN_INCLUDE_OBSERVE", False),
        min_score=(ps.get_float("PMS_PLAN_MIN_SCORE", 0.0) or None),
        min_sources=ps.get_int("PMS_PLAN_MIN_SOURCES", 0),
        min_upside=(ps.get_float("PMS_PLAN_MIN_UPSIDE", 0.0) or None),
        theme_cap=ps.get_int("PMS_PLAN_THEME_CAP_LOCAL", 5),
        exclude_st=ps.get_bool("PMS_PLAN_EXCLUDE_ST", True))
    print(f"[6] 按当前参数筛选 (top_n={ps.get_int('PMS_PLAN_TOP_N', 30)} "
          f"tiers={ps.get_list('PMS_PLAN_TIERS', [])} "
          f"observe={ps.get_bool('PMS_PLAN_INCLUDE_OBSERVE', False)} "
          f"PMS侧主题限额={ps.get_int('PMS_PLAN_THEME_CAP_LOCAL', 5)} "
          f"剔ST={ps.get_bool('PMS_PLAN_EXCLUDE_ST', True)})")
    tc = sel.get("tier_complete")
    tcs = ("是 —— 已看到白名单之外的档位, 后面再多也不影响候选池" if tc is True else
           ("**否** —— 返回的全是白名单档位, 说明被 top 卡住了, 调大 PMS_PLAN_TOP" if tc is False
            else "不适用 (没设档位白名单)"))
    print(f"    白名单档位取全: {tcs}")
    print(f"    排序池 {sel['considered']} → 合格 {sel['eligible']} → 取 {len(sel['items'])} 只")
    if sel.get("st_unknown"):
        print(f"    ! {len(sel['st_unknown'])} 只没有名称, ST 判不了已放行: {sel['st_unknown'][:8]}")
    print(f"    丢弃明细 {sel['dropped']}   (此处未扣持仓/黑名单, 下命令时还会再扣)")
    st = {}
    for x in sel["items"]:
        st[x.get("theme") or "(无)"] = st.get(x.get("theme") or "(无)", 0) + 1
    print("    入池主题分布: " + ", ".join(f"{k}×{v}" for k, v in
                                     sorted(st.items(), key=lambda y: -y[1])))
    print("    " + ", ".join(x["ts_code"] for x in sel["items"]))

    # --- [7] 榜单变化 (PMS 自算; 上游 changes 恒为 null, 见 UPSTREAM_PLAN_API §9)
    #     默认**只读**: 拿刚取到的这份跟库里最新快照比, 一个字都不写 —— 本脚本的契约
    #     是"只读探活", 破了它比少个功能糟得多。要落库请显式加 --snapshot。
    print("[7] 榜单变化" + (" (落库)" if args.snapshot else " (只读, 不写库)"))
    try:
        # **必须先比后写。** 反过来的话 preview 会拿这份计划跟"刚写进去的它自己"比,
        # 于是永远报"完全相同、无变化" —— 加个 --snapshot 就把要看的东西看没了。
        ch = pf.preview_changes(plan)
        if args.snapshot:
            snap = pf.snapshot(plan, keep=ps.get_int("PMS_PLAN_SNAPSHOT_KEEP", 200))
            print(f"    落快照: "
                  f"{'新版本已入库' if snap.get('stored') else '与库里最新那份相同, 未新增'} "
                  f"(digest={snap.get('digest', '')[:12]} {snap.get('rows')} 只)")
        d = ch.get("diff")
        if not ch.get("ok"):
            print(f"    {ch.get('hint') or '算不了'}")
        else:
            if d.get("prev_broken"):
                print("    ! 上一版名册**读不出来** (快照损坏) —— 这次比不了, "
                      "下面的空白不代表没有变化")
            elif ch.get("same_as_stored"):
                print("    与库里最新那份**完全相同** (指纹一致) —— 上游没重算")
            kind = {"first": "库里还没有快照, 无可比对的上一版 (跑一次 --snapshot 或页面强刷即有)",
                    "cross_day": "跨交易日的正常升降档",
                    "same_date_revision": "**同一计划日的新版本 —— 上游重算过** (§7.3 已知行为)"}
            print(f"    {kind.get(d['kind'], d['kind'])}")
            print(f"    {d['note']}")
            h = d.get("held") or {}
            for x in (h.get("exited") or []):
                print(f"    ! 持仓票掉榜: {x['ts_code']} {x['name']} "
                      f"(上版 rank {x['rank']} / {x['tier']})")
            for x in (h.get("tier_down") or []):
                print(f"    ! 持仓票降档: {x['ts_code']} {x['name']} "
                      f"{x['tier_from']} → {x['tier_to']}")
            for x in (h.get("coverage_lost") or []):
                print(f"    ! 持仓票丢了券商覆盖 (主榜→观察档): {x['ts_code']} {x['name']}")
            if not h.get("n_watch") and d["kind"] != "first" and not d.get("prev_broken"):
                print("    持仓票没被摘、没降档、也没丢券商覆盖")
        log = pf.snapshot_log(limit=30)
        if log.get("revised"):
            print("    同一计划日出现多版: "
                  + ", ".join(f"{k}×{v}" for k, v in log["revised"].items()))
    except Exception as e:
        print(f"    ! 算榜单变化失败: {type(e).__name__}: {e}")

    if args.json:
        print("[8] 解析结构")
        print(json.dumps({k: v for k, v in plan.items() if k not in ("main", "observe")},
                         ensure_ascii=False, indent=2))
    return 0


if __name__ == "__main__":
    sys.exit(main())