@gscdump/engine 0.19.0 → 0.19.1

package/dist/_chunks/engine.mjs

@@ -273,6 +273,23 @@ function normalizeRow(table, row) {
 		url: normalized
 	};
 }
+const DAILY_PARTITION_RE = /^daily\/(\d{4}-\d{2}-\d{2})$/;
+function queryRangeOf(partitions) {
+	if (!partitions) return void 0;
+	let min;
+	let max;
+	for (const p of partitions) {
+		const m = DAILY_PARTITION_RE.exec(p);
+		if (!m) continue;
+		const d = m[1];
+		if (min === void 0 || d < min) min = d;
+		if (max === void 0 || d > max) max = d;
+	}
+	return min !== void 0 ? {
+		start: min,
+		end: max
+	} : void 0;
+}
 function createStorageEngine(opts) {
 	const { dataSource, manifestStore, codec, executor } = opts;
 	const defaultNow = opts.now ?? (() => Date.now());
@@ -406,7 +423,7 @@ function createStorageEngine(opts) {
 				table: ref.table,
 				partitions: ref.partitions,
 				...opts.searchType !== void 0 ? { searchType: opts.searchType } : {}
-			})).map((e) => e.objectKey)];
+			}), queryRangeOf(ref.partitions)).map((e) => e.objectKey)];
 		}));
 		opts.signal?.throwIfAborted();
 		const fileKeys = {};

package/dist/_chunks/parquet-plan.mjs

@@ -186,9 +186,12 @@ function partitionSpan(partition) {
 		};
 	}
 }
-function splitOverlappingTiers(entries) {
+function splitOverlappingTiers(entries, queryRange) {
+	const rangeStartMs = queryRange ? Date.parse(`${queryRange.start}T00:00:00Z`) : void 0;
+	const rangeEndMs = queryRange ? Date.parse(`${queryRange.end}T00:00:00Z`) : void 0;
 	const spanned = [];
 	const kept = [];
+	const subsumed = [];
 	for (const entry of entries) {
 		const span = partitionSpan(entry.partition);
 		if (!span) {
@@ -196,7 +199,14 @@ function splitOverlappingTiers(entries) {
 			continue;
 		}
 		const days = [];
-		for (let t = span.startMs; t <= span.endMs; t += MS_PER_DAY) days.push(t);
+		for (let t = span.startMs; t <= span.endMs; t += MS_PER_DAY) {
+			if (rangeStartMs !== void 0 && (t < rangeStartMs || t > rangeEndMs)) continue;
+			days.push(t);
+		}
+		if (queryRange && days.length === 0) {
+			subsumed.push(entry);
+			continue;
+		}
 		spanned.push({
 			entry,
 			rank: span.rank,
@@ -205,7 +215,6 @@ function splitOverlappingTiers(entries) {
 	}
 	spanned.sort((a, b) => a.rank - b.rank || b.entry.createdAt - a.entry.createdAt);
 	const coveredBySearchType = /* @__PURE__ */ new Map();
-	const subsumed = [];
 	for (const { entry, days } of spanned) {
 		const slice = inferSearchType(entry);
 		let covered = coveredBySearchType.get(slice);
@@ -225,8 +234,8 @@ function splitOverlappingTiers(entries) {
 		subsumed
 	};
 }
-function dedupeOverlappingTiers(entries) {
-	return splitOverlappingTiers(entries).kept;
+function dedupeOverlappingTiers(entries, queryRange) {
+	return splitOverlappingTiers(entries, queryRange).kept;
 }
 function monthEndMs(month) {
 	const [y, m] = month.split("-").map(Number);

package/dist/_chunks/storage.d.mts

@@ -38,13 +38,27 @@ declare function enumeratePartitions(startDate: string, endDate: string): string
  * monthly) still double-counts those boundary days — eliminating that needs
  * per-file date predicates in the SQL, tracked separately. Unrecognised
  * partition shapes (`hourly/`, sidecar keys) are always kept.
+ *
+ * `queryRange` clamps every entry's day-span to the window the caller will
+ * actually read. This is required when `entries` came from a partition-
+ * filtered `listLive` (`runSQL` enumerates only the partitions intersecting
+ * the query): a `monthly/2026-04` whose Apr 27-30 falls past the query end
+ * must not be judged "unsubsumed" just because `weekly/2026-04-27` wasn't
+ * enumerated — those out-of-window days are SQL-filtered to nothing anyway.
+ * Omit `queryRange` when `entries` is the full manifest (e.g. analysis-sources).
  */
-declare function splitOverlappingTiers(entries: ManifestEntry[]): {
+declare function splitOverlappingTiers(entries: ManifestEntry[], queryRange?: {
+  start: string;
+  end: string;
+}): {
   kept: ManifestEntry[];
   subsumed: ManifestEntry[];
 };
 /** Entries worth reading — see {@link splitOverlappingTiers}. */
-declare function dedupeOverlappingTiers(entries: ManifestEntry[]): ManifestEntry[];
+declare function dedupeOverlappingTiers(entries: ManifestEntry[], queryRange?: {
+  start: string;
+  end: string;
+}): ManifestEntry[];
 /**
  * Default `searchType` for entries written before the field landed and for
  * sync paths that don't request a specific type. GSC's own default; the

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@gscdump/engine",
   "type": "module",
-  "version": "0.19.0",
+  "version": "0.19.1",
   "description": "Append-only Parquet/DuckDB storage engine + planner + adapters for the gscdump pipeline. Node + edge runtimes; opt-in heavy peers.",
   "author": {
     "name": "Harlan Wilton",
@@ -169,8 +169,8 @@
   "dependencies": {
     "drizzle-orm": "^0.45.2",
     "proper-lockfile": "^4.1.2",
-    "gscdump": "0.19.0",
-    "@gscdump/contracts": "0.19.0"
+    "gscdump": "0.19.1",
+    "@gscdump/contracts": "0.19.1"
   },
   "devDependencies": {
     "@duckdb/duckdb-wasm": "^1.32.0",