feat(cli): sync export — 日线按年导出 Parquet（data/parquet/stock_daily/<year>.parquet）

- 对应 ROADMAP §1.4「历史时序大数据转 Parquet」；pyarrow 随 qlib 依赖已可用
- 用法：uv run python -m app.cli.sync export [--years 2023,2024]
- 实测：9680 行 → 2023/2024 两个 parquet（data/parquet 已被 gitignore）
- ruff clean / pytest 86 passed

backend/app/cli/sync.py

@@ -169,6 +169,55 @@ def cmd_verify(args) -> int:
     return 0
 
 
+def cmd_export(args) -> int:
+    """把 SQLite 日线按年导出为 Parquet（data/parquet/stock_daily/<year>.parquet）。"""
+    from pathlib import Path
+
+    import pandas as pd
+
+    from app.infrastructure.persistence.sqlalchemy.models.market import StockDailyModel
+
+    settings = get_settings()
+    out_root = settings.storage.get("parquet_dir") or Path("data/parquet")
+    out_root.mkdir(parents=True, exist_ok=True)
+    total = 0
+    years = [int(y) for y in (args.years or "").split(",") if y.strip()] or None
+    with _session_ctx() as session:
+        all_bars = session.execute(
+            select(StockDailyModel).order_by(StockDailyModel.trade_date)
+        ).scalars()
+        frame = pd.DataFrame(
+            [
+                {
+                    "symbol": b.symbol,
+                    "trade_date": b.trade_date,
+                    "open": float(b.open) if b.open is not None else None,
+                    "high": float(b.high) if b.high is not None else None,
+                    "low": float(b.low) if b.low is not None else None,
+                    "close": float(b.close) if b.close is not None else None,
+                    "volume": float(b.volume) if b.volume is not None else None,
+                    "amount": float(b.amount) if b.amount is not None else None,
+                }
+                for b in all_bars
+            ]
+        )
+    if frame.empty:
+        print("[export] 无日线数据，请先运行 sync daily")
+        return 0
+    frame["trade_date"] = pd.to_datetime(frame["trade_date"])
+    out_dir = out_root / "stock_daily"
+    out_dir.mkdir(parents=True, exist_ok=True)
+    for year, group in frame.groupby(frame["trade_date"].dt.year):
+        if years and int(year) not in years:
+            continue
+        path = out_dir / f"{year}.parquet"
+        group.sort_values(["symbol", "trade_date"]).to_parquet(path, index=False)
+        total += len(group)
+        print(f"[export] {year} → {path}（{len(group)} 行）")
+    print(f"[export] 合计 {total} 行 → {out_dir}")
+    return 0
+
+
 def build_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(prog="app.cli.sync", description="Tushare 数据同步 CLI")
     sub = parser.add_subparsers(dest="command", required=True)
@@ -197,6 +246,10 @@ def build_parser() -> argparse.ArgumentParser:
     p_verify = sub.add_parser("verify", help="新浪交叉验证最新行情")
     p_verify.add_argument("--symbol", required=True)
     p_verify.set_defaults(func=cmd_verify)
+
+    p_export = sub.add_parser("export", help="日线按年导出 Parquet（data/parquet）")
+    p_export.add_argument("--years", default="", help="逗号分隔年份，留空导出全部")
+    p_export.set_defaults(func=cmd_export)
     return parser
 
 

docs/USAGE.md

@@ -108,6 +108,7 @@ uv run python -m app.cli.sync daily --symbols 600519.SH,000858.SZ --start 202301
 uv run python -m app.cli.sync daily --all --start 20240101 --resume   # 全市场 + 断点续传
 uv run python -m app.cli.sync financial --symbols 600519.SH
 uv run python -m app.cli.sync verify --symbol 600519.SH               # 新浪交叉验证
+uv run python -m app.cli.sync export --years 2024                # 日线按年导出 Parquet（默认全部）
 ```
 
 说明：