[reland] add ability to view multiple focused log lines at once

add ability to view multiple focused log lines at once

Add view of multiple lines using rockset. Currently we just show the
same error line (because there is only one in rockset). Currently, this
is also hidden under a feature flag.

<img width="915" alt="Screenshot 2023-10-13 at 2 28 46 PM"
src="https://github.com/pytorch/test-infra/assets/13758638/4652e132-6886-4c21-9ac4-f208cc4d1edd">

<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at a80550d</samp>

This pull request adds the log annotation feature to the test-infra web
app, which allows users to rate the quality of the log output for a job
and to see multiple failure lines for a job. It also updates the backend
SQL queries and the frontend components to handle the new failureLines
and failureLineNumbers properties for jobs, and fixes some minor issues
in the code. The affected files include `JobLinks.tsx`, `LogViewer.tsx`,
`LogAnnotationToggle.tsx`, `types.ts`, `drciUtils.ts`, `searchUtils.ts`,
`log_annotation/[repoOwner]/[repoName]/[annotation].ts`, `metrics.tsx`,
and several files in the `rockset` folder.

Original commit changeset: 3b5dd27

torchci/components/JobLinks.tsx

@@ -61,11 +61,11 @@ export default function JobLinks({ job }: { job: JobData }) {
       <TestInsightsLink job={job} separator={" | "} />
       <DisableTest job={job} label={"skipped"} />
       {authenticated && <UnstableJob job={job} label={"unstable"} />}
-      {authenticated && job.failureLine && (
+      {authenticated && job.failureLines && (
         <ReproductionCommand
           job={job}
           separator={" | "}
-          testName={getTestName(job.failureLine)}
+          testName={getTestName(job.failureLines[0])}
         />
       )}
     </span>
@@ -98,7 +98,7 @@ This test was disabled because it is failing on main branch ([recent examples]($
 }
 
 function DisableTest({ job, label }: { job: JobData; label: string }) {
-  const hasFailureClassification = job.failureLine != null;
+  const hasFailureClassification = job.failureLines != null;
   const swrKey = hasFailureClassification ? `/api/issue/${label}` : null;
   const { data } = useSWR(swrKey, fetcher, {
     // Set a 60s cache for the request, so that lots of tooltip hovers don't
@@ -114,7 +114,7 @@ function DisableTest({ job, label }: { job: JobData; label: string }) {
     return null;
   }
 
-  const testName = getTestName(job.failureLine!);
+  const testName = job.failureLines ? getTestName(job.failureLines[0]) : null;
   // - The failure classification is not a python unittest or pytest failure.
   if (testName === null) {
     return null;

torchci/components/LogViewer.tsx

@@ -194,13 +194,22 @@ function Log({ url, line }: { url: string; line: number | null }) {
 export default function LogViewer({
   job,
   logRating = LogAnnotation.NULL,
-  showAnnotationToggle = false,
+  showAnnotationToggle = process.env.DEBUG_LOG_CLASSIFIER === "true",
+  maxNumFailureLines = process.env.DEBUG_LOG_CLASSIFIER === "true" ? 2 : 1,
 }: {
   job: JobData;
   logRating?: LogAnnotation;
   showAnnotationToggle?: boolean;
+  maxNumFailureLines?: number;
 }) {
-  const [showLogViewer, setShowLogViewer] = useState(false);
+  // const numFailureLines =
+  //   Math.min(job.failureLines?.length || 0, maxNumFailureLines);
+
+  // we will replace this with the code above once we support having multiple failure lines in rockset
+  const numFailureLines = maxNumFailureLines;
+  const [showLogViewer, setShowLogViewer] = useState<boolean[]>(
+    Array.from({ length: numFailureLines }, () => false)
+  );
 
   useEffect(() => {
     document.addEventListener("copy", (e) => {
@@ -213,28 +222,39 @@ export default function LogViewer({
     });
   });
 
-  if (!job.failureLine && !isFailure(job.conclusion)) {
+  if (!job.failureLines && !isFailure(job.conclusion)) {
     return null;
   }
 
-  function handleClick() {
-    setShowLogViewer(!showLogViewer);
-  }
+  const toggleLogViewer = (index: number) => {
+    // Make a copy of the current array state
+    const updatedShowLogViewer = [...showLogViewer];
 
+    // Toggle the boolean value at the given index
+    updatedShowLogViewer[index] = !updatedShowLogViewer[index];
+
+    // Update the state
+    setShowLogViewer(updatedShowLogViewer);
+  };
   return (
     <div>
-      <button
-        style={{ background: "none", cursor: "pointer" }}
-        onClick={handleClick}
-      >
-        {showLogViewer ? "▼ " : "▶ "}
-        <code>{job.failureLine ?? "Show log"}</code>
-      </button>
-      {showLogViewer && <Log url={job.logUrl!} line={job.failureLineNumber!} />}
+      {showLogViewer.map((show, index) => (
+        <div key={index}>
+          <button
+            style={{ background: "none", cursor: "pointer" }}
+            onClick={() => toggleLogViewer(index)}
+          >
+            {show ? "▼ " : "▶ "}
+            <code>{job.failureLines ?? "Show log"}</code>
+          </button>
+          {show && <Log url={job.logUrl!} line={job.failureLineNumbers![0]} />}
+        </div>
+      ))}
       {showAnnotationToggle && (
         <div>
           <LogAnnotationToggle
             job={job}
+            // send in real metadata later
             log_metadata={{ job_id: "1" }}
             annotation={logRating}
           />

torchci/lib/drciUtils.ts

@@ -311,7 +311,7 @@ export async function hasSimilarFailures(
       head_sha: record.sha as string,
       head_branch: record.branch as string,
       failure_captures: record.failureCaptures as string[],
-      failure_line: record.failureLine,
+      failure_lines: record.failureLines,
     };
 
     // Only count different jobs with the same failure
@@ -333,9 +333,9 @@ export function isInfraFlakyJob(job: RecentWorkflowsData): boolean {
   // the workflow summary tab
   return (
     job.conclusion === "failure" &&
-    (job.failure_line === null ||
-      job.failure_line === undefined ||
-      job.failure_line === "") &&
+    (job.failure_lines === null ||
+      job.failure_lines === undefined ||
+      job.failure_lines.join("") === "") &&
     (job.runnerName === null ||
       job.runnerName === undefined ||
       job.runnerName === "")

torchci/lib/searchUtils.ts

@@ -95,10 +95,11 @@ export async function searchSimilarFailures(
       time: data.completed_at,
       conclusion: data.conclusion,
       htmlUrl: data.html_url,
-      failureLine: data.torchci_classification.line,
-      failureLineNumber: data.torchci_classification.line_num,
+      failureLines: [data.torchci_classification.line],
+      failureLineNumbers: [data.torchci_classification.line_num],
       failureCaptures: data.torchci_classification.captures,
     });
   });
+
   return { jobs: jobs };
 }

torchci/lib/types.ts

@@ -21,8 +21,8 @@ export interface JobData extends BasicJobData {
   logUrl?: string;
   durationS?: number;
   queueTimeS?: number;
-  failureLine?: string;
-  failureLineNumber?: number;
+  failureLines?: string[];
+  failureLineNumbers?: number[];
   failureCaptures?: string[];
   repo?: string;
   failureAnnotation?: string;
@@ -41,7 +41,7 @@ export interface RecentWorkflowsData extends BasicJobData {
   head_branch?: string | null;
   pr_number?: number;
   failure_captures: string[];
-  failure_line?: string | null;
+  failure_lines?: string[] | null;
 }
 
 export interface Artifact {

torchci/pages/api/drci/drci.ts

@@ -537,8 +537,8 @@ function isFlaky(job: RecentWorkflowsData, flakyRules: FlakyRule[]): boolean {
             failureCapture.match(captureRegex)
           );
         const matchFailureLine: boolean =
-          job.failure_line != null &&
-          job.failure_line.match(captureRegex) != null;
+          job.failure_lines != null &&
+          job.failure_lines[0].match(captureRegex) != null;
 
         // Accept both failure captures array and failure line string to make sure
         // that nothing is missing

torchci/pages/metrics.tsx

@@ -869,7 +869,6 @@ export default function Page() {
             additionalOptions={{ yAxis: { scale: true } }}
           />
         </Grid>
-
         <JobsDuration
           title={"Job time-to-signal, all branches"}
           branchName={"%"}

torchci/rockset/commons/__sql/annotated_flaky_jobs.sql

@@ -14,9 +14,9 @@ select
         PARSE_TIMESTAMP_ISO8601(job.completed_at)
     ) as durationS,
     w.repository.full_name as repo,
-    job.torchci_classification.line as failureLine,
+    ARRAY_CREATE(job.torchci_classification.line) as failureLines,
     job.torchci_classification.captures as failureCaptures,
-    job.torchci_classification.line_num as failureLineNumber,
+    ARRAY_CREATE(job.torchci_classification.line_num) as failureLineNumbers,
 from
     commons.job_annotation a
     join commons.workflow_job job on job.id = a.jobID