mirror of
https://ghfast.top/https://github.com/aeroxw/easy_tdx_max.git
synced 2026-09-12 23:54:21 +08:00
GitHub
9336273f17
新增导航页 /signals：把策略库 single/portfolio/multi 策略统一展开成 "策略×标的"子任务，按标的去重取最近 800 根 K 线，用与回测引擎同口径的 逐 bar 信号流程（含仓位跟踪）判断最近 N 根（窗口 1/3/5/10 可选，默认 5） 的买/卖信号，汇总卡片 + 筛选 tab + 明细表展示；上次结果缓存 localStorage。 - 后端 signal_scan.py（展开/去重取数/信号评估/汇总）+ POST /backtest/signal-scan/run/async - 只扫信号不重跑回测，不改写策略库业绩快照；单行失败（未知策略/停牌/参数非法）不中断整批 - normalize_symbol 按代码段纠正历史错标市场前缀（与前端 detectMarket 同规则） - 新增 20 个单测（含与回测引擎成交序列一致性对照），全套 1030 个单测通过
387 lines
15 KiB
Python
387 lines
15 KiB
Python
"""信号雷达（signal_scan）单元 + 端到端测试（离线，无网络）。
|
|
|
|
覆盖：
|
|
- normalize_symbol 市场前缀纠错
|
|
- expand_targets 三种 kind 展开 + 数据损坏容错
|
|
- fetch_scan_bars 去重取数 / 失败容错 / date→datetime 列归一化
|
|
- evaluate_signals 金叉买入、死叉卖出、仓位跟踪、窗口过滤（与回测引擎同口径）
|
|
- run_scan 单行失败不中断 + 汇总计数
|
|
- POST /backtest/signal-scan/run/async 端到端（fake store + fake 行情）
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import time
|
|
|
|
import numpy as np
|
|
import pandas as pd
|
|
import pytest
|
|
|
|
pytest.importorskip("fastapi")
|
|
|
|
from easy_tdx.web.signal_scan import ( # noqa: E402
|
|
evaluate_signals,
|
|
expand_targets,
|
|
fetch_scan_bars,
|
|
normalize_symbol,
|
|
run_scan,
|
|
)
|
|
from easy_tdx.web.strategy_store import SavedStrategy # noqa: E402
|
|
|
|
# ── 测试数据 ───────────────────────────────────────────────────────────────────
|
|
|
|
|
|
def v_shape_df(n_fall: int = 40, n_rise: int = 80, n_drop: int = 0) -> pd.DataFrame:
|
|
"""V 型走势合成日线：下跌 → 上涨（→ 可选急跌），保证出现金叉（→ 死叉）。
|
|
|
|
返回的 df 带标准 OHLCV + datetime 列（日线接口返回 date，归一化后是 datetime）。
|
|
"""
|
|
closes = np.concatenate(
|
|
[
|
|
10.0 - np.arange(n_fall) * 0.02, # 缓跌：MA5 持续低于 MA20
|
|
9.2 + np.arange(n_rise) * 0.12, # 稳定上涨：金叉出现
|
|
(10.0 + n_rise * 0.12 - np.arange(1, n_drop + 1) * 0.5) if n_drop else [], # 急跌：死叉
|
|
]
|
|
)
|
|
n = len(closes)
|
|
dates = pd.date_range("2025-01-01", periods=n, freq="B")
|
|
return pd.DataFrame(
|
|
{
|
|
"datetime": dates,
|
|
"open": closes - 0.05,
|
|
"high": closes + 0.10,
|
|
"low": closes - 0.10,
|
|
"close": closes,
|
|
"vol": np.full(n, 5000.0),
|
|
"amount": closes * 5000,
|
|
}
|
|
)
|
|
|
|
|
|
def _single(**ctx_overrides: object) -> SavedStrategy:
|
|
ctx: dict = {"symbol": "SH:601088", "category": "DAY"}
|
|
ctx.update(ctx_overrides)
|
|
return SavedStrategy(
|
|
id="s1",
|
|
name="神华·双均线",
|
|
kind="single",
|
|
strategy="ma_cross",
|
|
strategy_label="双均线交叉",
|
|
params={"fast": 5, "slow": 20},
|
|
context=ctx,
|
|
)
|
|
|
|
|
|
class FakeClient:
|
|
"""假行情客户端：按 (market, code) 返回预置 df，未预置的抛错。"""
|
|
|
|
def __init__(self, data: dict[str, pd.DataFrame]) -> None:
|
|
self.data = data
|
|
self.calls: list[tuple[str, str]] = []
|
|
|
|
async def get_security_bars(self, market, code, category, start, count): # noqa: ANN001
|
|
market_str = str(getattr(market, "name", market))
|
|
key = f"{market_str}:{code}"
|
|
self.calls.append((key, str(getattr(category, "name", category))))
|
|
if key not in self.data:
|
|
raise ConnectionError(f"no data for {key}")
|
|
return self.data[key]
|
|
|
|
|
|
# ── normalize_symbol ──────────────────────────────────────────────────────────
|
|
|
|
|
|
@pytest.mark.parametrize(
|
|
("raw", "expected"),
|
|
[
|
|
("SH:601088", "SH:601088"), # 正确的沪市主板
|
|
("SZ:515080", "SH:515080"), # 历史错标的沪市基金 → 纠正
|
|
("510300", "SH:510300"), # 无前缀 → 补全
|
|
("SZ:000001", "SZ:000001"), # 正确的深市主板
|
|
("430047", "BJ:430047"), # 北交所
|
|
("830799", "BJ:830799"), # 北交所 8xx
|
|
("SZ:300347", "SZ:300347"), # 创业板
|
|
],
|
|
)
|
|
def test_normalize_symbol(raw: str, expected: str) -> None:
|
|
assert normalize_symbol(raw) == expected
|
|
|
|
|
|
# ── expand_targets ────────────────────────────────────────────────────────────
|
|
|
|
|
|
def test_expand_single() -> None:
|
|
targets = expand_targets([_single()])
|
|
assert len(targets) == 1
|
|
t = targets[0]
|
|
assert (t.strategy, t.params, t.symbol, t.category) == (
|
|
"ma_cross",
|
|
{"fast": 5, "slow": 20},
|
|
"SH:601088",
|
|
"DAY",
|
|
)
|
|
assert t.error is None
|
|
|
|
|
|
def test_expand_portfolio_multi_symbols() -> None:
|
|
rec = SavedStrategy(
|
|
id="p1",
|
|
name="银行组合",
|
|
kind="portfolio",
|
|
strategy="macd",
|
|
params={"short": 10, "long": 20},
|
|
context={"stocks": ["SZ:000001", "515080", "SH:601088"], "category": "DAY"},
|
|
)
|
|
targets = expand_targets([rec])
|
|
assert len(targets) == 3
|
|
assert [t.symbol for t in targets] == ["SZ:000001", "SH:515080", "SH:601088"]
|
|
assert all(t.strategy == "macd" for t in targets)
|
|
|
|
|
|
def test_expand_multi_items() -> None:
|
|
rec = SavedStrategy(
|
|
id="m1",
|
|
name="老登+小登组合",
|
|
kind="multi",
|
|
strategy="multi",
|
|
context={
|
|
"items": [
|
|
{"strategy": "trix", "params": {"m1": 18, "m2": 20}, "symbol": "SZ:300347"},
|
|
{
|
|
"strategy": "ema_cross",
|
|
"params": {"fast": 12},
|
|
"symbol": "SZ:301308",
|
|
"category": "DAY",
|
|
},
|
|
{"strategy": "macd", "symbol": "SH:601088"}, # 缺 params → 默认空
|
|
{"strategy": "", "symbol": "SZ:000001"}, # 缺 strategy → error 行
|
|
]
|
|
},
|
|
)
|
|
targets = expand_targets([rec])
|
|
assert len(targets) == 4
|
|
ok = [t for t in targets if t.error is None]
|
|
assert [t.strategy for t in ok] == ["trix", "ema_cross", "macd"]
|
|
assert ok[1].params == {"fast": 12}
|
|
assert [t.error is None for t in targets] == [True, True, True, False]
|
|
|
|
|
|
def test_expand_error_rows() -> None:
|
|
# single 缺 symbol / multi 缺 items → 各展开为一条 error 行（不丢策略身份）
|
|
no_symbol = _single()
|
|
no_symbol.context = {"category": "DAY"}
|
|
broken_multi = SavedStrategy(id="m2", name="坏组合", kind="multi", strategy="multi")
|
|
targets = expand_targets([no_symbol, broken_multi])
|
|
assert len(targets) == 2
|
|
assert all(t.error for t in targets)
|
|
assert [t.strategy_name for t in targets] == ["神华·双均线", "坏组合"]
|
|
|
|
|
|
# ── fetch_scan_bars ───────────────────────────────────────────────────────────
|
|
|
|
|
|
def test_fetch_scan_bars_dedupe_and_normalize() -> None:
|
|
df = v_shape_df()
|
|
# 日线接口风格：date 列而非 datetime
|
|
daily = df.rename(columns={"datetime": "date"})
|
|
client = FakeClient({"SH:601088": daily, "SZ:000001": daily})
|
|
rec1 = _single()
|
|
rec2 = _single(id="s2", name="另一个神华", strategy="macd", params={})
|
|
targets = expand_targets([rec1, rec2]) # 同 symbol 只取一次
|
|
targets.append(expand_targets([_single(symbol="SZ:000001")])[0])
|
|
|
|
bars = asyncio.run(fetch_scan_bars(client, targets))
|
|
assert set(bars) == {("SH:601088", "DAY"), ("SZ:000001", "DAY")}
|
|
# SH:601088 只取了一次（去重生效）
|
|
assert len([c for c in client.calls if c[0] == "SH:601088"]) == 1
|
|
# date 列已归一化为 datetime 且按时间正序
|
|
out = bars[("SH:601088", "DAY")]
|
|
assert "datetime" in out.columns
|
|
assert out["datetime"].is_monotonic_increasing
|
|
|
|
|
|
def test_fetch_scan_bars_failure_tolerant() -> None:
|
|
client = FakeClient({}) # 全部抛错
|
|
targets = expand_targets([_single()])
|
|
bars = asyncio.run(fetch_scan_bars(client, targets))
|
|
assert bars == {("SH:601088", "DAY"): None}
|
|
|
|
|
|
# ── evaluate_signals ──────────────────────────────────────────────────────────
|
|
|
|
|
|
def _ma_cross_instance(): # noqa: ANN202
|
|
from easy_tdx.backtest.strategies import get_registry
|
|
|
|
return get_registry().get("ma_cross").build({"fast": 5, "slow": 20})
|
|
|
|
|
|
def _expected_cross_dates(df: pd.DataFrame, direction: str) -> list[str]:
|
|
"""用 MyTT 独立算出金叉/死叉所在日期（作为期望值，与被测代码解耦）。"""
|
|
from easy_tdx.MyTT import CROSS, MA
|
|
|
|
close = df["close"].to_numpy()
|
|
fast, slow = MA(close, 5), MA(close, 20)
|
|
mask = CROSS(fast, slow) if direction == "BUY" else CROSS(slow, fast)
|
|
return [str(df["datetime"].iloc[i])[:16] for i in range(len(df)) if mask[i]]
|
|
|
|
|
|
def test_evaluate_signals_golden_cross_buy() -> None:
|
|
df = v_shape_df(n_rise=30) # 只涨不跌：恰好一个金叉、之后无死叉
|
|
buy_dates = _expected_cross_dates(df, "BUY")
|
|
assert len(buy_dates) == 1, "V 型数据应恰好产生一个金叉"
|
|
cross_date = buy_dates[0]
|
|
cross_idx = [i for i in range(len(df)) if str(df["datetime"].iloc[i])[:16] == cross_date][0]
|
|
|
|
# 窗口恰好从金叉那根开始 → 窗口内能捕获 BUY
|
|
result = evaluate_signals(_ma_cross_instance(), df, window=len(df) - cross_idx)
|
|
buys = [s for s in result["recent_signals"] if s["direction"] == "BUY"]
|
|
assert [s["date"] for s in buys] == [cross_date]
|
|
assert result["latest_signal"] == "BUY"
|
|
assert result["signal_date"] == cross_date
|
|
assert result["position"] == "holding" # 买入后一直持有
|
|
assert result["last_close"] == pytest.approx(float(df["close"].iloc[-1]))
|
|
assert result["last_bar_date"] == str(df["datetime"].iloc[-1])[:16]
|
|
|
|
# 窗口再收窄一根（金叉在窗口外）→ 不上报旧信号，但仓位跟踪不受窗口影响
|
|
result2 = evaluate_signals(_ma_cross_instance(), df, window=len(df) - cross_idx - 1)
|
|
assert result2["recent_signals"] == []
|
|
assert result2["latest_signal"] is None
|
|
assert result2["position"] == "holding"
|
|
|
|
|
|
def test_evaluate_signals_death_cross_sell() -> None:
|
|
df = v_shape_df(n_drop=15) # 涨完急跌：金叉买入 → 死叉卖出
|
|
result = evaluate_signals(_ma_cross_instance(), df, window=len(df))
|
|
dirs = [s["direction"] for s in result["recent_signals"]]
|
|
assert dirs[0] == "BUY"
|
|
assert dirs[-1] == "SELL"
|
|
assert result["latest_signal"] == "SELL"
|
|
assert result["position"] == "flat" # 清仓
|
|
|
|
|
|
def test_evaluate_signals_matches_engine_trades() -> None:
|
|
"""与真实回测引擎成交方向序列一致性抽查（同 df、同策略）。"""
|
|
from easy_tdx.backtest.engine import BacktestEngine
|
|
|
|
df = v_shape_df(n_drop=15)
|
|
strat = _ma_cross_instance()
|
|
result = evaluate_signals(strat, df, window=len(df))
|
|
engine = BacktestEngine(strategy=_ma_cross_instance())
|
|
trades = engine.run(df).trades
|
|
engine_dirs = list(trades["direction"])
|
|
scan_dirs = [s["direction"] for s in result["recent_signals"]]
|
|
assert scan_dirs == engine_dirs[: len(scan_dirs)]
|
|
|
|
|
|
# ── run_scan ──────────────────────────────────────────────────────────────────
|
|
|
|
|
|
def test_run_scan_summary_and_errors() -> None:
|
|
df = v_shape_df()
|
|
targets = [
|
|
expand_targets([_single()])[0], # 正常行（有行情）
|
|
expand_targets([_single(id="s2", name="同标的第二策略")])[0], # 同标的复用行情
|
|
]
|
|
# 制造三类失败：未知策略 / 无行情 / 展开错误
|
|
bad_strategy = expand_targets([_single()])[0]
|
|
bad_strategy.strategy = "nope_strategy"
|
|
targets.append(bad_strategy)
|
|
no_bars = expand_targets([_single()])[0]
|
|
no_bars.symbol = "SZ:999999"
|
|
targets.append(no_bars)
|
|
broken = expand_targets([SavedStrategy(id="x", name="坏", kind="single", strategy="ma_cross")])[
|
|
0
|
|
]
|
|
targets.append(broken)
|
|
|
|
bars = {("SH:601088", "DAY"): df}
|
|
out = run_scan(bars, targets, window=len(df))
|
|
assert out["total"] == 5
|
|
assert out["buy_count"] == 2 # 前两行各有一个金叉买入
|
|
assert out["sell_count"] == 0
|
|
assert out["error_count"] == 3 # 未知策略 / 无行情 / 展开错误
|
|
rows = out["rows"]
|
|
assert rows[0]["error"] is None
|
|
assert rows[0]["latest_signal"] == "BUY"
|
|
assert "未知策略" in rows[2]["error"]
|
|
assert "未取到有效 K 线" in rows[3]["error"]
|
|
assert "缺少标的上下文" in rows[4]["error"]
|
|
assert out["elapsed"] >= 0
|
|
|
|
|
|
# ── API 端到端 ────────────────────────────────────────────────────────────────
|
|
|
|
|
|
@pytest.fixture()
|
|
def api_client():
|
|
from fastapi.testclient import TestClient
|
|
|
|
from easy_tdx.web import create_app
|
|
|
|
app = create_app()
|
|
with TestClient(app) as c:
|
|
yield c
|
|
|
|
|
|
def test_signal_scan_endpoint_e2e(api_client, monkeypatch) -> None:
|
|
"""POST 提交 → 轮询 done → 结果结构完整（fake store + fake 取数）。"""
|
|
import easy_tdx.web.signal_scan as sigscan
|
|
import easy_tdx.web.strategy_store as store_mod
|
|
|
|
# 缓跌 59 根 + 末根跳涨：金叉恰好发生在最后一根 K 线（窗口=1 也能捕获）
|
|
df = v_shape_df(n_fall=59, n_rise=0)
|
|
df.loc[df.index[-1], ["open", "high", "low", "close"]] = [14.95, 15.2, 14.8, 15.0]
|
|
|
|
class FakeStore:
|
|
def list_all(self) -> list[SavedStrategy]:
|
|
return [_single()]
|
|
|
|
async def fake_fetch(client, targets): # noqa: ANN001
|
|
return {("SH:601088", "DAY"): df}
|
|
|
|
monkeypatch.setattr(store_mod, "get_store", lambda: FakeStore())
|
|
monkeypatch.setattr(sigscan, "fetch_scan_bars", fake_fetch)
|
|
|
|
resp = api_client.post("/api/v1/backtest/signal-scan/run/async", json={"window_bars": 1})
|
|
assert resp.status_code == 202, resp.text
|
|
task_id = resp.json()["task_id"]
|
|
|
|
final = None
|
|
for _ in range(200):
|
|
poll = api_client.get(f"/api/v1/backtest/tasks/{task_id}")
|
|
assert poll.status_code == 200
|
|
final = poll.json()
|
|
if final["status"] in ("done", "failed"):
|
|
break
|
|
time.sleep(0.05)
|
|
assert final is not None and final["status"] == "done", final
|
|
|
|
result = final["result"]
|
|
assert result["total"] == 1
|
|
assert result["buy_count"] == 1
|
|
row = result["rows"][0]
|
|
assert row["strategy"] == "ma_cross"
|
|
assert row["symbol"] == "SH:601088"
|
|
assert row["error"] is None
|
|
assert row["position"] in ("holding", "flat")
|
|
|
|
|
|
def test_signal_scan_endpoint_empty_store(api_client, monkeypatch) -> None:
|
|
import easy_tdx.web.strategy_store as store_mod
|
|
|
|
class EmptyStore:
|
|
def list_all(self) -> list[SavedStrategy]:
|
|
return []
|
|
|
|
monkeypatch.setattr(store_mod, "get_store", lambda: EmptyStore())
|
|
resp = api_client.post("/api/v1/backtest/signal-scan/run/async", json={})
|
|
assert resp.status_code == 400
|
|
assert "策略库为空" in resp.json()["detail"]
|
|
|
|
|
|
def test_signal_scan_endpoint_window_validation(api_client) -> None:
|
|
resp = api_client.post("/api/v1/backtest/signal-scan/run/async", json={"window_bars": 0})
|
|
assert resp.status_code == 422
|