feat(HEL-435): 盘后未出数时晚间自动重试并提供安全补跑

Co-authored-by: multica-agent <github@multica.ai>

xiaobai-datahub/datahub/pipeline.py

@@ -22,6 +22,8 @@ LOGGER = get_logger()
 HARD_DATASETS = {"daily", "valuation", "index_daily"}
 SOFT_DATASETS = {"moneyflow", "auction"}
 OFFICIAL_DATASETS = HARD_DATASETS | SOFT_DATASETS
+EOD_A_DATASETS = ("daily", "valuation", "moneyflow", "auction")
+EOD_B_DATASETS = ("index_daily",)
 EMPTY_BATCH_ERROR = "empty official batch: 0 valid rows"
 
 STAGING_INSERT = {
@@ -379,14 +381,62 @@ class Pipeline:
             self._set_batch(batch_id, dataset, trade_date, "failed", 1, error=str(exc), finished=True)
             raise
 
+    def missing_official_datasets(self, trade_date: str) -> list[str]:
+        """Official datasets without an active publication for the date."""
+        day = yyyymmdd(trade_date)
+        placeholders = ",".join("?" for _ in OFFICIAL_DATASETS)
+        rows = self.db.fetchall(
+            f"SELECT dataset FROM publications WHERE trade_date = ? AND dataset IN ({placeholders})",
+            (day, *sorted(OFFICIAL_DATASETS)),
+        )
+        published = {str(row["dataset"]) for row in rows}
+        return [dataset for dataset in sorted(OFFICIAL_DATASETS) if dataset not in published]
+
+    def run_eod_missing(self, trade_date: str) -> dict[str, Any]:
+        """Fetch/publish every official dataset still missing for the date.
+
+        Idempotent: datasets with an existing publication are skipped, so
+        repeats never overwrite the current official batch. Per-dataset
+        failures are collected instead of aborting the remaining datasets.
+        """
+        return self._run_eod_datasets(tuple(sorted(OFFICIAL_DATASETS)), trade_date)
+
     def run_eod_batch_a(self, trade_date: str) -> dict[str, Any]:
-        results = {}
-        for dataset in ("daily", "valuation", "moneyflow", "auction"):
-            results[dataset] = self.run_dataset(dataset, trade_date)
-        return results
+        return self._run_eod_datasets(EOD_A_DATASETS, trade_date)
 
     def run_eod_batch_b(self, trade_date: str) -> dict[str, Any]:
-        return {"index_daily": self.run_dataset("index_daily", trade_date)}
+        return self._run_eod_datasets(EOD_B_DATASETS, trade_date)
+
+    def _run_eod_datasets(self, datasets: tuple[str, ...], trade_date: str) -> dict[str, Any]:
+        day = yyyymmdd(trade_date)
+        results: dict[str, Any] = {}
+        for dataset in datasets:
+            if self.active_batch(dataset, day) is not None:
+                results[dataset] = {
+                    "dataset": dataset,
+                    "trade_date": day,
+                    "state": "skipped",
+                    "reason": "already_published",
+                }
+                continue
+            try:
+                results[dataset] = self.run_dataset(dataset, day)
+            except Exception as exc:
+                results[dataset] = {
+                    "dataset": dataset,
+                    "trade_date": day,
+                    "state": "failed",
+                    "error": str(exc),
+                }
+        return results
+
+    @staticmethod
+    def eod_failures(results: dict[str, Any]) -> list[str]:
+        return [
+            f"{name}: {item.get('error')}"
+            for name, item in results.items()
+            if isinstance(item, dict) and item.get("state") == "failed"
+        ]
 
     def validate(self, dataset: str, batch_id: str, trade_date: str, rows: list[dict[str, Any]]) -> dict[str, Any]:
         quality = self.settings.quality