Logs: improve deduplication supporting code and fix duplicated frames (#88716)

* Logs: change refId approach to deduplicate infinite scrolling rows

* Chore: update tests

* chore: add comment

* Prettier

packages/grafana-o11y-ds-frontend/src/combineResponses.ts

@@ -159,9 +159,12 @@ function shouldCombine(frame1: DataFrame, frame2: DataFrame): boolean {
   // because we do not have a good "frametype" value for them yet.
   const customType1 = frame1.meta?.custom?.frameType;
   const customType2 = frame2.meta?.custom?.frameType;
-
+  // Legacy frames have this custom type
   if (customType1 === 'LabeledTimeValues' && customType2 === 'LabeledTimeValues') {
     return true;
+  } else if (customType1 === customType2) {
+    // Data plane frames don't
+    return true;
   }
 
   // should never reach here

public/app/features/explore/state/query.ts

@@ -37,7 +37,6 @@ import {
 } from 'app/core/utils/explore';
 import { getShiftedTimeRange } from 'app/core/utils/timePicker';
 import { getCorrelationsBySourceUIDs } from 'app/features/correlations/utils';
-import { infiniteScrollRefId } from 'app/features/logs/logsModel';
 import { getDatasourceSrv } from 'app/features/plugins/datasource_srv';
 import { getFiscalYearStartMonth, getTimeZone } from 'app/features/profile/state/selectors';
 import { SupportingQueryType } from 'app/plugins/datasource/loki/types';
@@ -57,7 +56,7 @@ import { notifyApp } from '../../../core/actions';
 import { createErrorNotification } from '../../../core/copy/appNotification';
 import { runRequest } from '../../query/state/runRequest';
 import { visualisationTypeKey } from '../Logs/utils/logs';
-import { decorateData } from '../utils/decorators';
+import { decorateData, decorateWithLogsResult } from '../utils/decorators';
 import {
   getSupplementaryQueryProvider,
   storeSupplementaryQueryEnabled,
@@ -583,8 +582,7 @@ export const runQueries = createAsyncThunk<void, RunQueriesOptions>(
           decorateData(
             data,
             queryResponse,
-            absoluteRange,
+            decorateWithLogsResult({ absoluteRange, refreshInterval, queries }),
-            refreshInterval,
             queries,
             correlations,
             showCorrelationEditorLinks,
@@ -646,8 +644,7 @@ export const runQueries = createAsyncThunk<void, RunQueriesOptions>(
           decorateData(
             data,
             queryResponse,
-            absoluteRange,
+            decorateWithLogsResult({ absoluteRange, refreshInterval, queries }),
-            refreshInterval,
             queries,
             correlations,
             showCorrelationEditorLinks,
@@ -754,7 +751,7 @@ export const runLoadMoreLogsQueries = createAsyncThunk<void, RunLoadMoreLogsQuer
       .map((query: DataQuery) => ({
         ...query,
         datasource: query.datasource || datasourceInstance?.getRef(),
-        refId: `${infiniteScrollRefId}${query.refId}`,
+        refId: query.refId,
         supportingQueryType: SupportingQueryType.InfiniteScroll,
       }));
 
@@ -791,8 +788,7 @@ export const runLoadMoreLogsQueries = createAsyncThunk<void, RunLoadMoreLogsQuer
           // This shouldn't be needed after https://github.com/grafana/grafana/issues/57327 is fixed
           combinePanelData(queryResponse, data),
           queryResponse,
-          absoluteRange,
+          decorateWithLogsResult({ absoluteRange, queries, deduplicate: true }),
-          undefined,
           logQueries,
           correlations,
           showCorrelationEditorLinks,

public/app/features/explore/utils/decorators.ts

@@ -264,6 +264,7 @@ export const decorateWithLogsResult =
       absoluteRange?: AbsoluteTimeRange;
       refreshInterval?: string;
       queries?: DataQuery[];
+      deduplicate?: boolean;
     } = {}
   ) =>
   (data: ExplorePanelData): ExplorePanelData => {
@@ -272,7 +273,13 @@ export const decorateWithLogsResult =
     }
 
     const intervalMs = data.request?.intervalMs;
-    const newResults = dataFrameToLogsModel(data.logsFrames, intervalMs, options.absoluteRange, options.queries);
+    const newResults = dataFrameToLogsModel(
+      data.logsFrames,
+      intervalMs,
+      options.absoluteRange,
+      options.queries,
+      options.deduplicate
+    );
     const sortOrder = refreshIntervalToSortOrder(options.refreshInterval);
     const sortedNewResults = sortLogsResult(newResults, sortOrder);
     const rows = sortedNewResults.rows;
@@ -286,8 +293,7 @@ export const decorateWithLogsResult =
 export function decorateData(
   data: PanelData,
   queryResponse: PanelData,
-  absoluteRange: AbsoluteTimeRange,
+  logsResultDecorator: (data: ExplorePanelData) => ExplorePanelData,
-  refreshInterval: string | undefined,
   queries: DataQuery[] | undefined,
   correlations: CorrelationData[] | undefined,
   showCorrelationEditorLinks: boolean,
@@ -305,7 +311,7 @@ export function decorateData(
     ),
     map(decorateWithFrameTypeMetadata),
     map(decorateWithGraphResult),
-    map(decorateWithLogsResult({ absoluteRange, refreshInterval, queries })),
+    map(logsResultDecorator),
     mergeMap(decorateWithRawPrometheusResult),
     mergeMap(decorateWithTableResult)
   );

public/app/features/logs/logsModel.test.ts

@@ -31,7 +31,6 @@ import {
   dedupLogRows,
   filterLogLevels,
   getSeriesProperties,
-  infiniteScrollRefId,
   LIMIT_LABEL,
   logRowToSingleRowDataFrame,
   logSeriesToLogsModel,
@@ -1023,12 +1022,12 @@ describe('dataFrameToLogsModel', () => {
     let frameA: DataFrame, frameB: DataFrame;
     beforeEach(() => {
       const { logFrameA, logFrameB } = getMockFrames();
-      logFrameA.refId = `${infiniteScrollRefId}-A`;
+      logFrameA.refId = `A`;
       logFrameA.fields[0].values = [1, 1];
       logFrameA.fields[1].values = ['line', 'line'];
       logFrameA.fields[3].values = ['3000000', '3000000'];
       logFrameA.fields[4].values = ['id', 'id'];
-      logFrameB.refId = `${infiniteScrollRefId}-B`;
+      logFrameB.refId = `B`;
       logFrameB.fields[0].values = [2, 2];
       logFrameB.fields[1].values = ['line 2', 'line 2'];
       logFrameB.fields[3].values = ['4000000', '4000000'];
@@ -1037,18 +1036,21 @@ describe('dataFrameToLogsModel', () => {
       frameB = logFrameB;
     });
 
-    it('deduplicates repeated log frames when invoked from infinite scrolling results', () => {
+    it('deduplicates repeated log frames when called with deduplicate', () => {
-      const logsModel = dataFrameToLogsModel([frameA, frameB], 1, { from: 1556270591353, to: 1556289770991 }, [
+      const logsModel = dataFrameToLogsModel(
-        { refId: `${infiniteScrollRefId}-A` },
+        [frameA, frameB],
-        { refId: `${infiniteScrollRefId}-B` },
+        1,
-      ]);
+        { from: 1556270591353, to: 1556289770991 },
+        [{ refId: `A` }, { refId: `B` }],
+        true
+      );
 
       expect(logsModel.rows).toHaveLength(2);
       expect(logsModel.rows[0].entry).toBe(frameA.fields[1].values[0]);
       expect(logsModel.rows[1].entry).toBe(frameB.fields[1].values[0]);
     });
 
-    it('does not remove repeated log frames when invoked from other contexts', () => {
+    it('does not remove repeated log frames when invoked without deduplicate', () => {
       frameA.refId = 'A';
       frameB.refId = 'B';
       const logsModel = dataFrameToLogsModel([frameA, frameB], 1, { from: 1556270591353, to: 1556289770991 }, [

public/app/features/logs/logsModel.ts

@@ -201,8 +201,6 @@ function isLogsData(series: DataFrame) {
   return series.fields.some((f) => f.type === FieldType.time) && series.fields.some((f) => f.type === FieldType.string);
 }
 
-export const infiniteScrollRefId = 'infinite-scroll-';
-
 /**
  * Convert dataFrame into LogsModel which consists of creating separate array of log rows and metrics series. Metrics
  * series can be either already included in the dataFrame or will be computed from the log rows.
@@ -215,29 +213,12 @@ export function dataFrameToLogsModel(
   dataFrame: DataFrame[],
   intervalMs?: number,
   absoluteRange?: AbsoluteTimeRange,
-  queries?: DataQuery[]
+  queries?: DataQuery[],
+  deduplicateResults?: boolean
 ): LogsModel {
-  // Until nanosecond precision for requests is supported, we need to account for possible duplicate rows.
-  let infiniteScrollingResults = false;
-  queries = queries?.map((query) => {
-    if (query.refId.includes(infiniteScrollRefId)) {
-      infiniteScrollingResults = true;
-      return {
-        ...query,
-        refId: query.refId.replace(infiniteScrollRefId, ''),
-      };
-    }
-    return query;
-  });
-  if (infiniteScrollingResults) {
-    dataFrame = dataFrame.map((frame) => ({
-      ...frame,
-      refId: frame.refId?.replace(infiniteScrollRefId, ''),
-    }));
-  }
-
   const { logSeries } = separateLogsAndMetrics(dataFrame);
-  const logsModel = logSeriesToLogsModel(logSeries, queries, infiniteScrollingResults);
+  // Until nanosecond precision for requests is supported, we need to account for possible duplicate rows.
+  const logsModel = logSeriesToLogsModel(logSeries, queries, Boolean(deduplicateResults));
 
   if (logsModel) {
     // Create histogram metrics from logs using the interval as bucket size for the line count