TT-7666 fix: PBT Transcribe shows duplicate audio after a segment boundary adjustment

src/renderer/src/components/PassageDetail/PassageDetailTranscribe.test.tsx

@@ -1,7 +1,20 @@
 import React from 'react';
 import { render, screen } from '@testing-library/react';
 
-let captured: { hasPermission?: boolean; curRole?: string } = {};
+let captured: {
+  hasPermission?: boolean;
+  curRole?: string;
+  phraseRegions?: unknown;
+} = {};
+
+/** Per-test knobs for the phrase-segment path (TT-7666). */
+const phrase = {
+  isPhraseArtifact: false,
+  slug: 'vernacular',
+  regions: [] as unknown[],
+  /** What `related()` answers for a row's sourceMedia. */
+  sourceMedia: undefined as string | undefined,
+};
 
 const linkedSharedResource = {
   id: 'sr1',
@@ -18,7 +31,10 @@ const passageDetailCtx = {
   orgWorkflowSteps: [
     {
       id: 'step-transcribe',
-      attributes: { sequencenum: 1, tool: '{"tool":"transcribe","settings":{}}' },
+      attributes: {
+        sequencenum: 1,
+        tool: '{"tool":"transcribe","settings":{}}',
+      },
     },
   ],
   setStepComplete: jest.fn(),
@@ -30,7 +46,10 @@ const passageDetailCtx = {
   sharedResource: undefined as unknown,
 };
 
-jest.mock('../../context/usePassageDetailContext', () => () => passageDetailCtx);
+jest.mock(
+  '../../context/usePassageDetailContext',
+  () => () => passageDetailCtx
+);
 
 jest.mock('../../context/PassageDetailContext', () => ({
   PassageDetailContext: React.createContext({ setState: jest.fn() }),
@@ -39,9 +58,11 @@ jest.mock('../../context/PassageDetailContext', () => ({
 jest.mock('../../context/TranscriberContext', () => ({
   TranscriberProvider: (props: {
     curRole?: string;
+    phraseRegions?: unknown;
     children?: React.ReactNode;
   }) => {
     captured.curRole = props.curRole;
+    captured.phraseRegions = props.phraseRegions;
     return <>{props.children}</>;
   },
 }));
@@ -83,20 +104,20 @@ jest.mock('../../crud', () => ({
 jest.mock('../../crud/useArtifactType', () => ({
   useArtifactType: () => ({
     localizedArtifactTypeFromId: () => 'bt',
-    slugFromId: () => 'vernacular',
+    slugFromId: () => phrase.slug,
   }),
 }));
 
 jest.mock('../../crud/artifactTypeSlug', () => ({
   ArtifactTypeSlug: { CarefulSpeech: 'carefulspeech' },
   artifactStampsStepLanguage: () => false,
-  isPhraseSegmentArtifact: () => false,
+  isPhraseSegmentArtifact: () => phrase.isPhraseArtifact,
 }));
 
 jest.mock('../../crud/related', () => ({
-  related: jest.fn(),
+  related: () => phrase.sourceMedia,
   __esModule: true,
-  default: jest.fn(),
+  default: () => phrase.sourceMedia,
 }));
 
 jest.mock('../../utils/useStepPermission', () => ({
@@ -106,7 +127,13 @@ jest.mock('../../utils/useStepPermission', () => ({
 }));
 
 jest.mock('../../hoc/useOrbitData', () => ({
-  useOrbitData: () => [],
+  useOrbitData: () => [
+    {
+      id: 'mf1',
+      type: 'mediafile',
+      attributes: { versionNumber: 1, segments: '[]' },
+    },
+  ],
 }));
 
 jest.mock('../../context/UnsavedContext', () => {
@@ -130,12 +157,12 @@ jest.mock('react-redux', () => ({
 
 jest.mock('../../utils/namedSegments', () => ({
   getSegments: () => '{}',
-  getSortedRegions: () => [],
+  getSortedRegions: () => phrase.regions,
   NamedRegions: { Clause: 'clause', BackTranslation: 'bt' },
 }));
 
 jest.mock('./carefulSpeech/carefulSpeechBoundary', () => ({
-  hasPhraseRegions: () => false,
+  hasPhraseRegions: () => phrase.regions.length > 0,
 }));
 
 jest.mock('./carefulSpeech/matchesGuidedOutputRow', () => ({
@@ -152,12 +179,19 @@ jest.mock('./boldClause/StepMessage', () => () => null);
 
 import { PassageDetailTranscribe } from './PassageDetailTranscribe';
 
+const resetKnobs = () => {
+  captured = {};
+  passageDetailCtx.sharedResource = undefined;
+  passageDetailCtx.mediafileId = 'mf1';
+  passageDetailCtx.rowData = [];
+  phrase.isPhraseArtifact = false;
+  phrase.slug = 'vernacular';
+  phrase.regions = [];
+  phrase.sourceMedia = undefined;
+};
+
 describe('PassageDetailTranscribe linked note (TT-5873)', () => {
-  beforeEach(() => {
-    captured = {};
-    passageDetailCtx.sharedResource = undefined;
-    passageDetailCtx.mediafileId = 'mf1';
-  });
+  beforeEach(resetKnobs);
 
   it('keeps transcribe permission on the source note', () => {
     render(<PassageDetailTranscribe width={400} artifactTypeId={null} />);
@@ -175,3 +209,38 @@ describe('PassageDetailTranscribe linked note (TT-5873)', () => {
     expect(captured.curRole).toBe('view');
   });
 });
+
+/**
+ * TT-7666 - the task list is built from every take attached to the vernacular,
+ * so the takes left behind by a segment-boundary adjustment showed up beside
+ * the ones recorded after it: two segments, four tasks to transcribe. Which
+ * takes are still current is decided against the segment boundaries the step is
+ * reading, so the provider has to be told what they are.
+ */
+describe('PassageDetailTranscribe phrase takes (TT-7666)', () => {
+  const clauseRegions = [
+    { start: 0, end: 6, label: '' },
+    { start: 6, end: 10, label: '' },
+  ];
+
+  beforeEach(() => {
+    resetKnobs();
+    phrase.sourceMedia = 'mf1';
+    passageDetailCtx.rowData = [
+      { artifactType: 'bt', mediafile: { id: 'take1', type: 'mediafile' } },
+    ] as never;
+  });
+
+  it('hands the current segment boundaries to the transcriber provider', () => {
+    phrase.isPhraseArtifact = true;
+    phrase.slug = 'carefulspeech';
+    phrase.regions = clauseRegions;
+    render(<PassageDetailTranscribe width={400} artifactTypeId={'art1'} />);
+    expect(captured.phraseRegions).toEqual(clauseRegions);
+  });
+
+  it('leaves a non-phrase artifact unscoped', () => {
+    render(<PassageDetailTranscribe width={400} artifactTypeId={'art1'} />);
+    expect(captured.phraseRegions ?? []).toEqual([]);
+  });
+});

src/renderer/src/components/PassageDetail/PassageDetailTranscribe.tsx

@@ -317,6 +317,7 @@ export function PassageDetailTranscribe({ width, artifactTypeId }: IProps) {
       artifactTypeId={artifactTypeId}
       curRole={curRole as string}
       stepLanguageBcp47={stepLanguageBcp47}
+      phraseRegions={phraseRegions}
     >
       <Grid
         container

src/renderer/src/components/PassageDetail/carefulSpeech/carefulSpeechCompletion.ts

@@ -5,8 +5,7 @@ import {
   matchesGuidedOutputRow,
   pickLatestGuidedOutputRow,
 } from './matchesGuidedOutputRow';
-
-const REGION_TOLERANCE = 0.05;
+import { parseTakeSourceRegion, regionsMatch } from '../../../crud/phraseTakes';
 
 function isEmptySourceSegments(seg: string | undefined): boolean {
   if (!seg) return true;
@@ -20,19 +19,6 @@ function isEmptySourceSegments(seg: string | undefined): boolean {
   }
 }
 
-function parseStoredRegion(seg: string | undefined): IRegion | undefined {
-  if (!seg) return undefined;
-  try {
-    const parsed = JSON.parse(seg) as IRegion;
-    if (parsed?.start !== undefined && parsed?.end !== undefined) {
-      return parsed;
-    }
-  } catch {
-    return undefined;
-  }
-  return undefined;
-}
-
 function regionMatchesClause(
   storedSeg: string | undefined,
   clauseRegion: IRegion,
@@ -45,13 +31,8 @@ function regionMatchesClause(
   ) {
     return true;
   }
-  const stored = parseStoredRegion(storedSeg);
-  if (stored) {
-    return (
-      Math.abs(stored.start - clauseRegion.start) < REGION_TOLERANCE &&
-      Math.abs(stored.end - clauseRegion.end) < REGION_TOLERANCE
-    );
-  }
+  const stored = parseTakeSourceRegion(storedSeg);
+  if (stored) return regionsMatch(stored, clauseRegion);
   return prettySegment(storedSeg).trim() === prettySegment(clauseRegion).trim();
 }
 

src/renderer/src/components/PassageDetail/carefulSpeech/matchesGuidedOutputRow.ts

@@ -1,6 +1,7 @@
 import { related } from '../../../crud/related';
 import { IRow } from '../../../context/PassageDetailContext';
 import { mediaMatchesStepLanguage } from '../../../utils/mediaLanguage';
+import { compareTakesNewestFirst } from '../../../crud/phraseTakes';
 
 // Language-field helpers live in utils/mediaLanguage so context-layer callers
 // don't have to import from this component subtree. Re-exported here because
@@ -47,12 +48,9 @@ export function matchesGuidedOutputRow(
 export function pickLatestGuidedOutputRow(matches: IRow[]): IRow | undefined {
   if (matches.length === 0) return undefined;
   if (matches.length === 1) return matches[0];
-  return [...matches].sort((a, b) => {
-    const da = a.mediafile?.attributes?.dateCreated ?? '';
-    const db = b.mediafile?.attributes?.dateCreated ?? '';
-    if (da !== db) return db.localeCompare(da);
-    return (b.mediafile?.id ?? '').localeCompare(a.mediafile?.id ?? '');
-  })[0];
+  return [...matches].sort((a, b) =>
+    compareTakesNewestFirst(a.mediafile, b.mediafile)
+  )[0];
 }
 
 /** Named-region key for Phrase BT segment boundaries for a language. */

src/renderer/src/context/TranscriberContext.tsx

@@ -33,6 +33,8 @@ import {
 } from '../crud';
 import { mediaFileName } from '../crud/media';
 import { mediaMatchesStepLanguage } from '../utils/mediaLanguage';
+import { selectCurrentPhraseTakes } from '../crud/phraseTakes';
+import { IRegion } from '../crud/useWavesurferRegions';
 import StickyRedirect from '../components/StickyRedirect';
 import { useSelector } from 'react-redux';
 import { useDispatch } from 'react-redux';
@@ -133,9 +135,15 @@ interface IProps {
   curRole?: string;
   /** Step language. When set (and not `und`), only media tagged with it become tasks. */
   stepLanguageBcp47?: string;
+  /**
+   * Phrase-segment boundaries the step is reading. When given, a segment
+   * contributes its newest take and the takes left behind by a boundary
+   * adjustment are not tasks (TT-7666).
+   */
+  phraseRegions?: IRegion[];
 }
 const TranscriberProvider = (props: IProps) => {
-  const { artifactTypeId, curRole, stepLanguageBcp47 } = props;
+  const { artifactTypeId, curRole, stepLanguageBcp47, phraseRegions } = props;
   const [isDetail] = useState(artifactTypeId !== undefined);
   const passages = useOrbitData<Passage[]>('passage');
   const sections = useOrbitData<Section[]>('section');
@@ -200,10 +208,19 @@ const TranscriberProvider = (props: IProps) => {
       return;
     }
     m = m.filter((mf) => mediaMatchesStepLanguage(mf, stepLanguageBcp47));
+    m = selectCurrentPhraseTakes(m, phraseRegions ?? []);
     setPlanMedia(m);
     planMediaRef.current = m;
     // eslint-disable-next-line react-hooks/exhaustive-deps
-  }, [mediafiles, devPlan, artifactId, stepLanguageBcp47, pasId, memory]);
+  }, [
+    mediafiles,
+    devPlan,
+    artifactId,
+    stepLanguageBcp47,
+    phraseRegions,
+    pasId,
+    memory,
+  ]);
 
   const setRows = (rowData: IRowData[]) => {
     setState((state: ICtxState) => {