claude_dev
59da65b839
两个根因:
1. relative_metrics 缺字段(routes.py): get_result 的 relative_fields 列表
漏了 total_return/annual_return/sharpe_ratio/max_drawdown 4 个字段, 导致
这 4 个指标永远进不了响应 → 前端 MetricCards 显"—"。补全为 11 字段。
2. 任务 task_id 不持久(runner/cta_engine): submit_cta 用 cta_{symbol}_{id(params)}
(内存地址) 作 runner id, 而 run_cta_backtest 内部另生成 cta_{uuid} 存 DB,
两者持久层不相交 → 重启后 pool 内存映射丢失, get_result(runner_id) 的
load_result_by_task_id 查不到 → /result 404 → 前端指标全 0 + 图表无数据。
改为 submit 前置生成稳定 uuid, 透传给 run_cta_backtest 复用, 使
runner-id == DB task_id (单一 id, 重启可查)。
附: Dashboard 最近任务"策略/因子"列 min-width 150→190(长策略名不再截断)
测试: test_runner task_id 断言同步新格式
187 lines
7.7 KiB
Python
187 lines
7.7 KiB
Python
"""
|
|
Orchestrator for task coordination and execution
|
|
Manages backtesting tasks with lazy imports
|
|
"""
|
|
import asyncio
|
|
import uuid
|
|
from concurrent.futures import Future
|
|
from .pool import TaskPool
|
|
from .task import TaskState
|
|
|
|
|
|
class Orchestrator:
|
|
"""Task coordinator for backtesting operations"""
|
|
|
|
def __init__(self, db_path: str, file_dir=None, max_workers: int = 2):
|
|
"""Initialize orchestrator with database path and worker limits"""
|
|
self.db_path = db_path
|
|
self.file_dir = file_dir
|
|
self.pool = TaskPool(max_workers=max_workers)
|
|
self._pending: dict[str, dict] = {}
|
|
self._on_stage = None # async callback(task_id, stage)
|
|
|
|
def set_on_stage(self, cb):
|
|
"""Set callback for stage updates (async callable)"""
|
|
self._on_stage = cb
|
|
|
|
async def _notify_stage(self, task_id: str, stage: str) -> None:
|
|
"""Update task stage and fire callback if set"""
|
|
self.pool.update_stage(task_id, stage)
|
|
if self._on_stage:
|
|
await self._on_stage(task_id, stage)
|
|
|
|
async def submit_cta(self, strategy_class, symbol: str, params: dict,
|
|
start: str, end: str, cfg, benchmark: str = "hs300") -> str:
|
|
"""Submit a CTA backtesting task asynchronously"""
|
|
# Stable uuid up front → reused as the persisted DB task_id, so runner-id ==
|
|
# DB task_id (durable across restarts; previously used id(params) memory addr).
|
|
task_id = f"cta_{uuid.uuid4().hex[:8]}"
|
|
self.pool.submit(task_id, "cta")
|
|
self._pending[task_id] = dict(
|
|
strategy_class=strategy_class,
|
|
symbol=symbol,
|
|
params=params,
|
|
start=start,
|
|
end=end,
|
|
cfg=cfg,
|
|
benchmark=benchmark
|
|
)
|
|
await self._notify_stage(task_id, "排队中")
|
|
|
|
spec = self._pending[task_id]
|
|
fut: Future = self.pool.submit_work(
|
|
task_id, _cta_worker, spec["strategy_class"], spec["symbol"],
|
|
spec["params"], spec["start"], spec["end"], spec["cfg"], spec["benchmark"], self.db_path, task_id
|
|
)
|
|
|
|
task = self.pool.get_task(task_id)
|
|
task.start()
|
|
await self._notify_stage(task_id, "回测中")
|
|
asyncio.ensure_future(self._wait_future(task_id, fut))
|
|
return task_id
|
|
|
|
async def submit_optimize(self, strategy_class, symbol: str, grid: dict,
|
|
start: str, end: str, cfg) -> str:
|
|
"""Submit a CTA optimization task asynchronously"""
|
|
task_id = f"opt_{symbol}_{id(grid)}"
|
|
self.pool.submit(task_id, "optimize")
|
|
self._pending[task_id] = dict(
|
|
strategy_class=strategy_class,
|
|
symbol=symbol,
|
|
grid=grid,
|
|
start=start,
|
|
end=end,
|
|
cfg=cfg
|
|
)
|
|
await self._notify_stage(task_id, "参数优化中")
|
|
|
|
spec = self._pending[task_id]
|
|
fut: Future = self.pool.submit_work(
|
|
task_id, _opt_worker, spec["strategy_class"], spec["symbol"],
|
|
spec["grid"], spec["start"], spec["end"], spec["cfg"], self.db_path
|
|
)
|
|
|
|
task = self.pool.get_task(task_id)
|
|
task.start()
|
|
await self._notify_stage(task_id, "参数优化中")
|
|
asyncio.ensure_future(self._wait_future(task_id, fut))
|
|
return task_id
|
|
|
|
async def submit_factor(self, symbols: list, factor_names: list,
|
|
start: str, end: str, cfg, output_dir: str) -> str:
|
|
"""Submit a factor analysis task asynchronously"""
|
|
task_id = f"factor_{id(factor_names)}"
|
|
self.pool.submit(task_id, "factor")
|
|
self._pending[task_id] = dict(
|
|
symbols=symbols,
|
|
factor_names=factor_names,
|
|
start=start,
|
|
end=end,
|
|
cfg=cfg,
|
|
output_dir=output_dir
|
|
)
|
|
await self._notify_stage(task_id, "因子分析中")
|
|
|
|
spec = self._pending[task_id]
|
|
fut: Future = self.pool.submit_work(
|
|
task_id, _factor_worker, spec["symbols"], spec["factor_names"],
|
|
spec["start"], spec["end"], spec["cfg"], spec["output_dir"]
|
|
)
|
|
|
|
task = self.pool.get_task(task_id)
|
|
task.start()
|
|
await self._notify_stage(task_id, "分析中")
|
|
asyncio.ensure_future(self._wait_future(task_id, fut))
|
|
return task_id
|
|
|
|
async def _wait_future(self, task_id: str, fut: Future) -> None:
|
|
"""Wait for Future to complete and handle result/exception
|
|
|
|
Bridges concurrent.futures.Future (from ProcessPoolExecutor) to asyncio coroutine.
|
|
"""
|
|
try:
|
|
result = await asyncio.wrap_future(fut)
|
|
await self._on_done(task_id, result)
|
|
except Exception as e:
|
|
task = self.pool.get_task(task_id)
|
|
if task:
|
|
task.fail(f"{type(e).__name__}: {e}")
|
|
await self._notify_stage(task_id, "失败")
|
|
|
|
async def _on_done(self, task_id: str, result) -> None:
|
|
"""Handle task completion (with None-guard for unknown tasks)"""
|
|
task = self.pool.get_task(task_id)
|
|
if task is None:
|
|
# Unknown task - fire callback but don't crash
|
|
await self._notify_stage(task_id, "完成")
|
|
return
|
|
|
|
# S1.1: use the persisted DB row id (BacktestResult.id) so get_result can
|
|
# load_result(result.id). FactorReport (no .id) falls back to None until S2.
|
|
task.complete(result_id=getattr(result, "id", None))
|
|
task.raw_result = result # S2: keep in-memory result (FactorReport) for ic-summary/report
|
|
await self._notify_stage(task_id, "完成")
|
|
|
|
def get_status(self, task_id: str) -> TaskState | None:
|
|
"""Get task status by ID"""
|
|
return self.pool.get_status(task_id)
|
|
|
|
def get_result(self, task_id: str):
|
|
"""Get task result by ID. Tries in-memory (current run) then DB (history)."""
|
|
task = self.pool.get_task(task_id)
|
|
if task and task.status == TaskState.DONE and task.result_id:
|
|
# Lazy import to avoid vnpy dependency issues
|
|
from sanguo_backtest.result_store import load_result
|
|
return load_result(task.result_id, self.db_path)
|
|
# Fallback: historical task persisted in DB (e.g. after restart)
|
|
from sanguo_backtest.result_store import load_result_by_task_id
|
|
return load_result_by_task_id(task_id, self.db_path)
|
|
|
|
def get_raw_result(self, task_id: str):
|
|
"""Get the raw in-memory result object (e.g. FactorReport) by task ID.
|
|
|
|
Used by factor endpoints (ic-summary, tears report) where the result
|
|
isn't a BacktestResult persisted to the DB.
|
|
"""
|
|
task = self.pool.get_task(task_id)
|
|
return task.raw_result if task else None
|
|
|
|
|
|
# Module-level worker functions (must be top-level for ProcessPoolExecutor pickle)
|
|
def _cta_worker(strategy_class, symbol: str, params: dict, start: str, end: str, cfg, benchmark: str, db_path: str, task_id: str) -> any:
|
|
"""Worker for CTA backtest (lazy import, spawn-friendly)"""
|
|
from sanguo_backtest.cta_engine import run_cta_backtest
|
|
return run_cta_backtest(strategy_class, symbol, params, start, end, cfg, db_path, benchmark=benchmark, task_id=task_id)
|
|
|
|
|
|
def _opt_worker(strategy_class, symbol: str, grid: dict, start: str, end: str, cfg, db_path: str) -> any:
|
|
"""Worker for CTA optimization (lazy import, spawn-friendly)"""
|
|
from sanguo_backtest.cta_optimizer import run_cta_optimization
|
|
return run_cta_optimization(strategy_class, symbol, grid, start, end, cfg, db_path)
|
|
|
|
|
|
def _factor_worker(symbols: list, factor_names: list, start: str, end: str, cfg, output_dir: str) -> any:
|
|
"""Worker for factor analysis (lazy import, spawn-friendly)"""
|
|
from sanguo_factor.analyzer import run_factor_analysis
|
|
return run_factor_analysis(symbols, factor_names, start, end, cfg, output_dir)
|