Merge pull request #16138 from calixteman/bug1820909

Fix search in pdf a containing some UTF-32 characters (bug 1820909)

test/pdfs/.gitignore

@@ -576,3 +576,4 @@
 !bug1770750.pdf
 !issue16063.pdf
 !issue16067.pdf
+!bug1820909.1.pdf

test/pdfs/bug1820909.pdf.link

@@ -0,0 +1,2 @@
+https://web.archive.org/web/20221122204959/https://www.unicode.org/charts/PDF/U31350.pdf
+

test/test_manifest.json

@@ -7455,5 +7455,12 @@
       "rounds": 1,
       "link": true,
       "type": "eq"
+   },
+   {
+      "id": "bug1820909",
+      "file": "pdfs/bug1820909.pdf",
+      "md5": "d95a83a868671a03cbf322f16b2e2b9d",
+      "link": true,
+      "type": "other"
    }
 ]

test/unit/pdf_find_controller_spec.js

@@ -854,4 +854,50 @@ describe("pdf_find_controller", function () {
       pageMatchesLength: [[7]],
     });
   });
+
+  it("performs a search in a text with some UTF-32 chars", async function () {
+    if (isNodeJS) {
+      pending("Linked test-cases are not supported in Node.js.");
+    }
+
+    const { eventBus, pdfFindController } = await initPdfFindController(
+      "bug1820909.pdf"
+    );
+
+    await testSearch({
+      eventBus,
+      pdfFindController,
+      state: {
+        query: "31350",
+      },
+      matchesPerPage: [1, 2],
+      selectedMatch: {
+        pageIndex: 0,
+        matchIndex: 0,
+      },
+      pageMatches: [[41], [131, 1359]],
+      pageMatchesLength: [[5], [5, 5]],
+    });
+  });
+
+  it("performs a search in a text with some UTF-32 chars followed by a dash at the end of a line", async function () {
+    const { eventBus, pdfFindController } = await initPdfFindController(
+      "bug1820909.1.pdf"
+    );
+
+    await testSearch({
+      eventBus,
+      pdfFindController,
+      state: {
+        query: "abcde",
+      },
+      matchesPerPage: [2],
+      selectedMatch: {
+        pageIndex: 0,
+        matchIndex: 0,
+      },
+      pageMatches: [[42, 95]],
+      pageMatchesLength: [[5, 5]],
+    });
+  });
 });

web/pdf_find_controller.js

@@ -289,21 +289,26 @@ function normalize(text) {
         // "X-\n" is removed because an hyphen at the end of a line
         // with not a space before is likely here to mark a break
         // in a word.
-        // The \n isn't in the original text so here y = i, n = 1 and o = 2.
-        positions.push([i - shift + 1, 1 + shift]);
+        // If X is encoded with UTF-32 then it can have a length greater than 1.
+        // The \n isn't in the original text so here y = i, n = X.len - 2 and
+        // o = X.len - 1.
+        const len = p5.length - 2;
+        positions.push([i - shift + len, 1 + shift]);
         shift += 1;
         shiftOrigin += 1;
         eol += 1;
-        return p5.charAt(0);
+        return p5.slice(0, -2);
       }
 
       if (p6) {
         // An ideographic at the end of a line doesn't imply adding an extra
         // white space.
-        positions.push([i - shift + 1, shift]);
+        // A CJK can be encoded in UTF-32, hence their length isn't always 1.
+        const len = p6.length - 1;
+        positions.push([i - shift + len, shift]);
         shiftOrigin += 1;
         eol += 1;
-        return p6.charAt(0);
+        return p6.slice(0, -1);
       }
 
       if (p7) {