Merge pull request #14483 from calixteman/200B

Remove the invisible format marks from the text chunks

src/core/evaluator.js

@@ -2561,6 +2561,9 @@ class PartialEvaluator {
 
       for (let i = 0, ii = glyphs.length; i < ii; i++) {
         const glyph = glyphs[i];
+        if (glyph.isInvisibleFormatMark) {
+          continue;
+        }
         let charSpacing =
           textState.charSpacing + (i + 1 === ii ? extraSpacing : 0);
 
@@ -2601,7 +2604,7 @@ class PartialEvaluator {
         // Must be called after compareWithLastPosition because
         // the textContentItem could have been flushed.
         const textChunk = ensureTextContentItem();
-        if (glyph.isDiacritic) {
+        if (glyph.isZeroWidthDiacritic) {
           scaledDim = 0;
         }
 

src/core/fonts.js

@@ -216,7 +216,8 @@ class Glyph {
 
     const category = getCharUnicodeCategory(unicode);
     this.isWhitespace = category.isWhitespace;
-    this.isDiacritic = category.isDiacritic;
+    this.isZeroWidthDiacritic = category.isZeroWidthDiacritic;
+    this.isInvisibleFormatMark = category.isInvisibleFormatMark;
   }
 
   matchesForCache(

src/core/unicode.js

@@ -1640,12 +1640,13 @@ function reverseIfRtl(chars) {
   return buf.join("");
 }
 
-const SpecialCharRegExp = new RegExp("^(\\s)|(\\p{Mn})$", "u");
+const SpecialCharRegExp = new RegExp("^(\\s)|(\\p{Mn})|(\\p{Cf})$", "u");
 function getCharUnicodeCategory(char) {
   const groups = char.match(SpecialCharRegExp);
   return {
     isWhitespace: !!(groups && groups[1]),
-    isDiacritic: !!(groups && groups[2]),
+    isZeroWidthDiacritic: !!(groups && groups[2]),
+    isInvisibleFormatMark: !!(groups && groups[3]),
   };
 }
 

test/pdfs/issue9186.pdf.link

@@ -0,0 +1 @@
+https://github.com/mozilla/pdf.js/files/1500985/Sample.lease.contract.26.pdf

test/test_manifest.json

@@ -6230,5 +6230,13 @@
       "firstPage": 1,
       "lastPage": 1,
       "type": "eq"
+   },
+   { "id": "issue9186",
+      "file": "pdfs/issue9186.pdf",
+      "md5": "d151857bb724ab9a291704a45a0b5d7f",
+      "rounds": 1,
+      "link": true,
+      "lastPage": 1,
+      "type": "text"
    }
 ]

test/unit/api_spec.js

@@ -2147,6 +2147,29 @@ sozialökonomische Gerechtigkeit.`)
       await loadingTask.destroy();
     });
 
+    it("gets text content, with invisible text marks (issue 9186)", async function () {
+      if (isNodeJS) {
+        pending("Linked test-cases are not supported in Node.js.");
+      }
+
+      const loadingTask = getDocument(buildGetDocumentParams("issue9186.pdf"));
+      const pdfDoc = await loadingTask.promise;
+      const pdfPage = await pdfDoc.getPage(1);
+      const { items } = await pdfPage.getTextContent();
+      const text = mergeText(items);
+
+      expect(
+        text.includes(`This Agreement (“Agreement”) is made as of this 25th day of January, 2017, by and
+between EDWARD G. ATSINGER III, not individually but as sole Trustee of the ATSINGER
+FAMILY TRUST /u/a dated October 31, 1980 as amended, and STUART W. EPPERSON, not
+individually but solely as Trustee of the STUART W. EPPERSON REVOCABLE LIVING
+TRUST /u/a dated January 14th 1993 as amended, collectively referred to herein as “Lessor”, and
+Caron Broadcasting, Inc., an Ohio corporation (“Lessee”).`)
+      ).toEqual(true);
+
+      await loadingTask.destroy();
+    });
+
     it("gets text content, with beginbfrange operator handled correctly (bug 1627427)", async function () {
       const loadingTask = getDocument(
         buildGetDocumentParams("bug1627427_reduced.pdf")

test/unit/unicode_spec.js

@@ -47,19 +47,67 @@ describe("unicode", function () {
     it("should correctly determine the character category", function () {
       const tests = {
         // Whitespace
-        " ": { isDiacritic: false, isWhitespace: true },
-        "\t": { isDiacritic: false, isWhitespace: true },
-        "\u2001": { isDiacritic: false, isWhitespace: true },
-        "\uFEFF": { isDiacritic: false, isWhitespace: true },
+        " ": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: true,
+        },
+        "\t": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: true,
+        },
+        "\u2001": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: true,
+        },
+        "\uFEFF": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: true,
+        },
 
         // Diacritic
-        "\u0302": { isDiacritic: true, isWhitespace: false },
-        "\u0344": { isDiacritic: true, isWhitespace: false },
-        "\u0361": { isDiacritic: true, isWhitespace: false },
+        "\u0302": {
+          isZeroWidthDiacritic: true,
+          isInvisibleFormatMark: false,
+          isWhitespace: false,
+        },
+        "\u0344": {
+          isZeroWidthDiacritic: true,
+          isInvisibleFormatMark: false,
+          isWhitespace: false,
+        },
+        "\u0361": {
+          isZeroWidthDiacritic: true,
+          isInvisibleFormatMark: false,
+          isWhitespace: false,
+        },
 
-        // No whitespace or diacritic
-        a: { isDiacritic: false, isWhitespace: false },
-        1: { isDiacritic: false, isWhitespace: false },
+        // Invisible format mark
+        "\u200B": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: true,
+          isWhitespace: false,
+        },
+        "\u200D": {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: true,
+          isWhitespace: false,
+        },
+
+        // No whitespace or diacritic or invisible format mark
+        a: {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: false,
+        },
+        1: {
+          isZeroWidthDiacritic: false,
+          isInvisibleFormatMark: false,
+          isWhitespace: false,
+        },
       };
       for (const [character, expectation] of Object.entries(tests)) {
         expect(getCharUnicodeCategory(character)).toEqual(expectation);