@gscdump/analysis 0.18.2 → 0.18.3

package/dist/analyzer/index.mjs

@@ -891,7 +891,10 @@ const changePointAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         ${valueExpr} AS value
@@ -1579,7 +1582,10 @@ const ctrAnomalyAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS day_clicks,
         ${METRIC_EXPR.impressions} AS day_impressions,
         ${METRIC_EXPR.ctr} AS day_ctr,
@@ -3533,7 +3539,10 @@ const positionVolatilityAnalyzer = defineAnalyzer({
       SELECT
         url AS page,
         query,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.impressions} AS q_impressions,
         ${METRIC_EXPR.position} AS q_position
       FROM read_parquet({{FILES}}, union_by_name = true)
@@ -3899,7 +3908,7 @@ const seasonalityAnalyzer = defineAnalyzer({
 			sql: `
       WITH monthly AS (
         SELECT
-          strftime(date, '%Y-%m') AS month,
+          strftime(CAST(date AS DATE), '%Y-%m') AS month,
           CAST(SUM(${params.metric === "impressions" ? "impressions" : "clicks"}) AS DOUBLE) AS value
         FROM read_parquet({{FILES}}, union_by_name = true)
         WHERE date >= ? AND date <= ?
@@ -3998,7 +4007,10 @@ const stlDecomposeAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         CAST(SUM(${metric}) AS DOUBLE) AS observed

package/dist/default-registry.mjs

@@ -891,7 +891,10 @@ const changePointAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         ${valueExpr} AS value
@@ -1579,7 +1582,10 @@ const ctrAnomalyAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS day_clicks,
         ${METRIC_EXPR.impressions} AS day_impressions,
         ${METRIC_EXPR.ctr} AS day_ctr,
@@ -3527,7 +3533,10 @@ const positionVolatilityAnalyzer = defineAnalyzer({
       SELECT
         url AS page,
         query,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.impressions} AS q_impressions,
         ${METRIC_EXPR.position} AS q_position
       FROM read_parquet({{FILES}}, union_by_name = true)
@@ -3893,7 +3902,7 @@ const seasonalityAnalyzer = defineAnalyzer({
 			sql: `
       WITH monthly AS (
         SELECT
-          strftime(date, '%Y-%m') AS month,
+          strftime(CAST(date AS DATE), '%Y-%m') AS month,
           CAST(SUM(${params.metric === "impressions" ? "impressions" : "clicks"}) AS DOUBLE) AS value
         FROM read_parquet({{FILES}}, union_by_name = true)
         WHERE date >= ? AND date <= ?
@@ -3992,7 +4001,10 @@ const stlDecomposeAnalyzer = defineAnalyzer({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         CAST(SUM(${metric}) AS DOUBLE) AS observed

package/dist/index.d.mts

@@ -467,6 +467,19 @@ declare function diffSitemapHealth(prev: readonly SitemapHealthRow[], curr: read
 interface SyncedRange {
   oldestDateSynced: string | null;
   newestDateSynced: string | null;
+  /**
+   * Optional sorted list of `[start, end]` daily-key spans (`YYYY-MM-DD`,
+   * both inclusive) that the engine actually has partitions for. When set,
+   * `shouldRouteToLive` returns true for any requested range that overlaps
+   * a day NOT inside one of these spans — even when the request sits inside
+   * `oldestDateSynced..newestDateSynced`. Lets the composite catch *internal*
+   * manifest gaps (e.g. a missing monthly tier) that the outer envelope
+   * doesn't reveal. Spans must be sorted by `start` and non-overlapping.
+   */
+  coveredSpans?: ReadonlyArray<{
+    start: string;
+    end: string;
+  }>;
 }
 interface CompositeSourceOptions {
   engine: AnalysisQuerySource$1;

package/dist/index.mjs

@@ -1107,7 +1107,10 @@ const changePointAnalyzer = defineAnalyzer$1({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         ${valueExpr} AS value
@@ -1795,7 +1798,10 @@ const ctrAnomalyAnalyzer = defineAnalyzer$1({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS day_clicks,
         ${METRIC_EXPR.impressions} AS day_impressions,
         ${METRIC_EXPR.ctr} AS day_ctr,
@@ -3888,7 +3894,10 @@ const positionVolatilityAnalyzer = defineAnalyzer$1({
       SELECT
         url AS page,
         query,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.impressions} AS q_impressions,
         ${METRIC_EXPR.position} AS q_position
       FROM read_parquet({{FILES}}, union_by_name = true)
@@ -4254,7 +4263,7 @@ const seasonalityAnalyzer = defineAnalyzer$1({
 			sql: `
       WITH monthly AS (
         SELECT
-          strftime(date, '%Y-%m') AS month,
+          strftime(CAST(date AS DATE), '%Y-%m') AS month,
           CAST(SUM(${params.metric === "impressions" ? "impressions" : "clicks"}) AS DOUBLE) AS value
         FROM read_parquet({{FILES}}, union_by_name = true)
         WHERE date >= ? AND date <= ?
@@ -4353,7 +4362,10 @@ const stlDecomposeAnalyzer = defineAnalyzer$1({
       SELECT
         query,
         url AS page,
-        date,
+        -- Normalize at the source CTE: union_by_name=true can coerce date to
+        -- VARCHAR across parquets with mixed schemas, which makes downstream
+        -- strftime(date, ...) binder-error.
+        CAST(date AS DATE) AS date,
         ${METRIC_EXPR.clicks} AS clicks,
         ${METRIC_EXPR.impressions} AS impressions,
         CAST(SUM(${metric}) AS DOUBLE) AS observed
@@ -6738,12 +6750,29 @@ function diffSitemapHealth(prev, curr) {
 		totals
 	};
 }
+function hasGapInCoveredSpans(start, end, coveredSpans) {
+	let cursor = start;
+	for (const span of coveredSpans) {
+		if (span.end < cursor) continue;
+		if (span.start > cursor) return true;
+		if (span.end >= end) return false;
+		cursor = nextDay(span.end);
+		if (cursor > end) return false;
+	}
+	return cursor <= end;
+}
+function nextDay(day) {
+	const t = Date.parse(`${day}T00:00:00Z`) + 864e5;
+	return new Date(t).toISOString().slice(0, 10);
+}
 function shouldRouteToLive(state, site) {
 	if (!canProxyToGsc(state)) return false;
 	const { startDate, endDate } = extractDateRange(state.filter);
 	if (!startDate || !endDate) return false;
 	if (!site.oldestDateSynced || !site.newestDateSynced) return true;
-	return startDate < site.oldestDateSynced || endDate > site.newestDateSynced;
+	if (startDate < site.oldestDateSynced || endDate > site.newestDateSynced) return true;
+	if (site.coveredSpans && site.coveredSpans.length > 0) return hasGapInCoveredSpans(startDate, endDate, site.coveredSpans);
+	return false;
 }
 function createCompositeSource(opts) {
 	const { engine, live, site } = opts;

package/dist/source/index.d.mts

@@ -4,6 +4,19 @@ import { PlannerCapabilities } from "gscdump/query/plan";
 interface SyncedRange {
   oldestDateSynced: string | null;
   newestDateSynced: string | null;
+  /**
+   * Optional sorted list of `[start, end]` daily-key spans (`YYYY-MM-DD`,
+   * both inclusive) that the engine actually has partitions for. When set,
+   * `shouldRouteToLive` returns true for any requested range that overlaps
+   * a day NOT inside one of these spans — even when the request sits inside
+   * `oldestDateSynced..newestDateSynced`. Lets the composite catch *internal*
+   * manifest gaps (e.g. a missing monthly tier) that the outer envelope
+   * doesn't reveal. Spans must be sorted by `start` and non-overlapping.
+   */
+  coveredSpans?: ReadonlyArray<{
+    start: string;
+    end: string;
+  }>;
 }
 interface CompositeSourceOptions {
   engine: AnalysisQuerySource;

package/dist/source/index.mjs

@@ -1,11 +1,28 @@
 import { canProxyToGsc } from "@gscdump/engine-gsc-api";
 import { extractDateRange } from "gscdump/query";
+function hasGapInCoveredSpans(start, end, coveredSpans) {
+	let cursor = start;
+	for (const span of coveredSpans) {
+		if (span.end < cursor) continue;
+		if (span.start > cursor) return true;
+		if (span.end >= end) return false;
+		cursor = nextDay(span.end);
+		if (cursor > end) return false;
+	}
+	return cursor <= end;
+}
+function nextDay(day) {
+	const t = Date.parse(`${day}T00:00:00Z`) + 864e5;
+	return new Date(t).toISOString().slice(0, 10);
+}
 function shouldRouteToLive(state, site) {
 	if (!canProxyToGsc(state)) return false;
 	const { startDate, endDate } = extractDateRange(state.filter);
 	if (!startDate || !endDate) return false;
 	if (!site.oldestDateSynced || !site.newestDateSynced) return true;
-	return startDate < site.oldestDateSynced || endDate > site.newestDateSynced;
+	if (startDate < site.oldestDateSynced || endDate > site.newestDateSynced) return true;
+	if (site.coveredSpans && site.coveredSpans.length > 0) return hasGapInCoveredSpans(startDate, endDate, site.coveredSpans);
+	return false;
 }
 function createCompositeSource(opts) {
 	const { engine, live, site } = opts;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@gscdump/analysis",
   "type": "module",
-  "version": "0.18.2",
+  "version": "0.18.3",
   "description": "GSC analyzers — striking-distance, opportunity, movers, decay, brand, clustering, concentration, seasonality. Pure row-based + DuckDB-native.",
   "author": {
     "name": "Harlan Wilton",
@@ -75,9 +75,9 @@
   },
   "dependencies": {
     "drizzle-orm": "^0.45.2",
-    "@gscdump/engine": "0.18.2",
-    "@gscdump/engine-gsc-api": "0.18.2",
-    "gscdump": "0.18.2"
+    "@gscdump/engine": "0.18.3",
+    "@gscdump/engine-gsc-api": "0.18.3",
+    "gscdump": "0.18.3"
   },
   "devDependencies": {
     "vitest": "^4.1.6"