fix(web): keep synthetic fingerprints out of the errors sparkline query

apps/web/src/components/errors/errors-hub.tsx

@@ -20,7 +20,12 @@ import {
 } from "@/lib/services/atoms/warehouse-query-atoms"
 import { retainedQueryV2 } from "@/lib/services/common/v2-atom-client"
 import { errorIssueFromV2 } from "@/lib/services/error-issues"
-import { buildErrorSignals, indexInvestigationsByIssue, type ErrorSignal } from "@/lib/models/error-signal"
+import {
+	buildErrorSignals,
+	indexInvestigationsByIssue,
+	sparkFingerprintHashes,
+	type ErrorSignal,
+} from "@/lib/models/error-signal"
 import {
 	clearedSelection,
 	type IssueSelectionMsg,
@@ -277,7 +282,7 @@ export function ErrorsHub(props: ErrorsHubProps) {
 
 	// The sparkline set follows the rows actually on screen, so a filtered view
 	// never pays for fingerprints it will not draw.
-	const fingerprintHashes = useMemo(() => issues.map((issue) => issue.fingerprintHash), [issues])
+	const fingerprintHashes = useMemo(() => sparkFingerprintHashes(issues), [issues])
 
 	const bucketSeconds = useMemo(() => {
 		const startMs = Date.parse(props.range.startTime.replace(" ", "T") + "Z")

apps/web/src/lib/models/error-signal.test.ts

@@ -3,7 +3,7 @@ import { describe, expect, it } from "vitest"
 import type { WorkflowState } from "@maple/domain/http"
 import type { V2Investigation } from "@maple/domain/http/v2"
 
-import { densifySpark, resolveSignalState, surgeRatio } from "./error-signal"
+import { densifySpark, resolveSignalState, sparkFingerprintHashes, surgeRatio } from "./error-signal"
 
 const HOUR_MS = 3_600_000
 
@@ -48,6 +48,21 @@ describe("resolveSignalState", () => {
 	})
 })
 
+describe("sparkFingerprintHashes", () => {
+	it("keeps only the issues whose fingerprint the warehouse can parse", () => {
+		// Alert and integration issues reuse the fingerprint column for a
+		// synthetic key. Sending one to the batched spark query makes ClickHouse
+		// reject `toUInt64('alert:…')` and fails the sparklines for every row.
+		const hashes = sparkFingerprintHashes([
+			{ kind: "error", fingerprintHash: "753390793895937054" },
+			{ kind: "alert", fingerprintHash: "alert:550bb2c5-8fe0-4ab9-b2f8-e566674cb0a2:scraper" },
+			{ kind: "integration", fingerprintHash: "planetscale:maple:branch.out_of_memory" },
+			{ kind: "error", fingerprintHash: "1037926342141719040" },
+		])
+		expect(hashes).toEqual(["753390793895937054", "1037926342141719040"])
+	})
+})
+
 describe("densifySpark", () => {
 	const window = { startMs: 0, endMs: 4 * HOUR_MS, bucketMs: HOUR_MS }
 

apps/web/src/lib/models/error-signal.ts

@@ -10,7 +10,13 @@
 //
 // An `ErrorSignal` is those three joined on the fingerprint hash.
 
-import type { ErrorIssueDocument, ErrorIssueId, IssueSeverity, WorkflowState } from "@maple/domain/http"
+import type {
+	ErrorIssueDocument,
+	ErrorIssueId,
+	IssueKind,
+	IssueSeverity,
+	WorkflowState,
+} from "@maple/domain/http"
 import type { V2Investigation } from "@maple/domain/http/v2"
 
 import type { ErrorByType } from "@/api/warehouse/errors"
@@ -126,6 +132,22 @@ export function resolveSignalState(
 	return { kind: "workflow", state: issue.workflowState }
 }
 
+/**
+ * The fingerprints a sparkline can actually be drawn for.
+ *
+ * Only `kind === "error"` issues carry a real ClickHouse fingerprint (a decimal
+ * UInt64 string). Alert and integration issues reuse the column for a synthetic
+ * key — `alert:{ruleId}:{groupKey}`, `planetscale:{database}:{event}` — which
+ * `toUInt64()` rejects, failing the whole batched spark query for every row on
+ * screen. `ErrorIssueReadModelsService.getIssue` skips the warehouse for the
+ * same reason.
+ */
+export function sparkFingerprintHashes(
+	issues: ReadonlyArray<{ readonly kind: IssueKind; readonly fingerprintHash: string }>,
+): ReadonlyArray<string> {
+	return issues.filter((issue) => issue.kind === "error").map((issue) => issue.fingerprintHash)
+}
+
 /**
  * Join issues (the spine) with warehouse volume and trend.
  *