fix: fix pdf text box mapping

packages/discovery-react-components/src/components/DocumentPreview/components/PdfHighlight/utils/textBoxMapping/MappingSourceTextProvider.ts

@@ -76,4 +76,11 @@ export class MappingSourceTextProvider {
   isBlank(text: string): boolean {
     return this.normalizer.isBlank(text);
   }
+
+  /**
+   * Cleanup the consumed text position history
+   */
+  resetHistory() {
+    this.provider.resetHistory();
+  }
 }

packages/discovery-react-components/src/components/DocumentPreview/components/PdfHighlight/utils/textBoxMapping/TextProvider.ts

@@ -133,4 +133,11 @@ export class TextProvider {
     });
     this.history = validSpans.slice(0, MAX_HISTORY);
   }
+
+  /**
+   * Cleanup the consumed text position history
+   */
+  resetHistory() {
+    this.history = [];
+  }
 }

packages/discovery-react-components/src/components/DocumentPreview/components/PdfHighlight/utils/textBoxMapping/__tests__/getTextBoxMapping.test.ts

@@ -61,4 +61,35 @@ describe('getTextBoxMapping', () => {
     expect(normalized2?.cell).toBe(target.cellAt(1)); // mapped to 2nd cell in target
     expect(normalized2?.span).toEqual([0, 3]); // mapped to the span [0, 3] in the 2nd cell
   });
+
+  it('should map text after long match properly', () => {
+    const source = new SimpleTextLayout([
+      // text_mappings box
+      {
+        text: '1 abc def ghi jkl 2 abc def ghi jkl 3 abc def ghi jkl 4 abc def ghi jkl',
+        bbox: bbox(0, 0, 2, 4)
+      }
+    ]);
+    const target = new SimpleTextLayout([
+      // 1st line
+      { bbox: bbox(0, 0), text: '1 abc def ghi' },
+      { bbox: bbox(1, 0), text: 'jkl' },
+      // 2nd line
+      { bbox: bbox(0, 1), text: '2 abc def ghi' },
+      { bbox: bbox(1, 1), text: 'jkl' },
+      // 3nd line
+      { bbox: bbox(0, 2), text: '3 abc def ghi' },
+      { bbox: bbox(1, 2), text: 'jkl' },
+      // 4th line
+      { bbox: bbox(0, 3), text: '4 abc def ghi' },
+      { bbox: bbox(1, 3), text: 'jkl' }
+    ]);
+    const mapping = getTextBoxMappings(source, target);
+
+    // verify the mapping result of the 1st 'jkl' in the source
+    const result1 = mapping.apply(source.cellAt(0).getPartial([14, 17]));
+    expect(result1).toHaveLength(1);
+    const normalized1 = result1[0].cell?.getNormalized();
+    expect(normalized1?.cell).toBe(target.cellAt(1)); // mapped to first cell in target
+  });
 });

packages/discovery-react-components/src/components/DocumentPreview/components/PdfHighlight/utils/textBoxMapping/getTextBoxMapping.ts

@@ -51,6 +51,7 @@ export function getTextBoxMappings<
         matchInSource.markAsMapped();
       }
     });
+    source.resetHistory();
   }
   return builder.toTextBoxMapping();
 }
@@ -178,6 +179,13 @@ class Source<SourceCell extends TextLayoutCell, TargetCell extends TextLayoutCel
     };
   }
 
+  /**
+   * Cleanup history in sources to process text from the start
+   */
+  resetHistory() {
+    this.sourceProviders.forEach(provider => provider.resetHistory());
+  }
+
   /**
    * Find the best source (larger text layout cell) where text `textToMatch` is in
    * @param sources source (larger) text layout cells overlapping the current target cell

packages/discovery-react-components/src/components/DocumentPreview/components/PdfHighlight/utils/textLayout/PdfTextContentTextLayout.ts

@@ -1,5 +1,6 @@
 import { bboxesIntersect } from 'components/DocumentPreview/utils/box';
 import { PDFPageViewport, PDFPageViewportOptions, TextContentItem } from 'pdfjs-dist';
+import { sortBy } from 'lodash';
 import { Bbox, TextSpan } from '../../types';
 import { BaseTextLayoutCell } from './BaseTextLayout';
 import { getAdjustedCellByOffsetByDom } from './dom';
@@ -18,16 +19,29 @@ export class PdfTextContentTextLayout implements TextLayout<PdfTextContentTextLa
 
     const textContentItems = textContentInfo.textContent.items;
 
-    this.cells = textContentItems.map((item, index) => {
+    const htmlBboxIndicesByIndex = textContentItems.map(item => {
       const cellBbox = getBbox(item, this.viewport);
-      let isInHtmlBbox = false;
       if (htmlBboxInfo?.bboxes?.length) {
-        isInHtmlBbox = htmlBboxInfo.bboxes.some(bbox => {
+        return htmlBboxInfo.bboxes.findIndex(bbox => {
           return bboxesIntersect(cellBbox, [bbox.left, bbox.top, bbox.right, bbox.bottom]);
         });
       }
+      return -1;
+    });
+    const cells = textContentItems.map((item, index) => {
+      const cellBbox = getBbox(item, this.viewport);
+      const isInHtmlBbox = htmlBboxIndicesByIndex[index] >= 0;
       return new PdfTextContentTextLayoutCell(this, index, item, pageNum, cellBbox, isInHtmlBbox);
     });
+
+    const htmlBboxIndicesPatched = htmlBboxIndicesByIndex.reduce(
+      (acc, value, index) => {
+        acc[index] = value >= 0 ? value : index > 0 ? acc[index - 1] : -1;
+        return acc;
+      },
+      [...htmlBboxIndicesByIndex]
+    );
+    this.cells = sortBy(cells, cell => htmlBboxIndicesPatched[cell.id]);
   }
 
   /**

yarn.lock

@@ -2163,7 +2163,7 @@ __metadata:
   languageName: node
   linkType: hard
 
-"@ibm-watson/discovery-react-components@^1.5.0-beta.29, @ibm-watson/discovery-react-components@workspace:packages/discovery-react-components":
+"@ibm-watson/discovery-react-components@^1.5.0-beta.30, @ibm-watson/discovery-react-components@workspace:packages/discovery-react-components":
   version: 0.0.0-use.local
   resolution: "@ibm-watson/discovery-react-components@workspace:packages/discovery-react-components"
   dependencies:
@@ -10664,7 +10664,7 @@ __metadata:
   dependencies:
     "@carbon/icons": ^10.5.0
     "@cypress/webpack-preprocessor": ^5.11.0
-    "@ibm-watson/discovery-react-components": ^1.5.0-beta.29
+    "@ibm-watson/discovery-react-components": ^1.5.0-beta.30
     "@ibm-watson/discovery-styles": ^1.5.0-beta.24
     "@testing-library/cypress": ^7.0.7
     "@types/proper-url-join": ^2