Fix issue #1013

src/UglyToad.PdfPig.Tests/Integration/GithubIssuesTests.cs

@@ -7,6 +7,29 @@
 
     public class GithubIssuesTests
     {
+        [Fact]
+        public void Issue1013()
+        {
+            // NB: We actually do not fix issue 953 here, but another bug found with the same document.
+            var path = IntegrationHelpers.GetSpecificTestDocumentPath("document_with_failed_fonts.pdf");
+
+            // Lenient parsing ON + Skip missing fonts
+            using (var document = PdfDocument.Open(path, new ParsingOptions() { UseLenientParsing = true, SkipMissingFonts = true }))
+            {
+                var page2 = document.GetPage(2);
+                Assert.NotEmpty(page2.Letters);
+
+                var words2 = NearestNeighbourWordExtractor.Instance.GetWords(page2.Letters).ToArray();
+                Assert.Equal("Doplňující", words2[0].Text);
+
+                var page3 = document.GetPage(3);
+                Assert.NotEmpty(page3.Letters);
+
+                var words3 = NearestNeighbourWordExtractor.Instance.GetWords(page3.Letters).ToArray();
+                Assert.Equal("Vinohradská", words3[8].Text);
+            }
+        }
+
         [Fact]
         public void Issue1016()
         {

src/UglyToad.PdfPig/Tokenization/Scanner/PdfTokenScanner.cs

@@ -835,11 +835,13 @@
             }
 
             if (!stream.StreamDictionary.TryGet(NameToken.First, out var firstToken)
-            || !(firstToken is NumericToken))
+            || !(firstToken is NumericToken firstTokenNum))
             {
                 throw new PdfDocumentFormatException($"Object stream dictionary did not provide first object offset {stream.StreamDictionary}.");
             }
 
+            long firstTokenOffset = firstTokenNum.Long;
+
             // Read the N integers
             var bytes = new MemoryInputBytes(stream.Decode(filterProvider, this));
 
@@ -854,7 +856,7 @@
                 scanner.MoveNext();
                 var byteOffset = (NumericToken)scanner.CurrentToken;
 
-                objects.Add((objectNumber.Long, byteOffset.Long));
+                objects.Add((objectNumber.Long, firstTokenOffset + byteOffset.Long));
             }
 
             var results = new List<ObjectToken>();
@@ -863,6 +865,16 @@
             {
                 var obj = objects[i];
 
+                // Check item offset is in [currentPosition - 1; currentPosition + 1]
+                bool isBetween = ((obj.Item2 - (scanner.CurrentPosition - 1)) | ((scanner.CurrentPosition + 1) - obj.Item2)) >= 0;
+                if (!isBetween)
+                {
+                    // TODO - Not sure if it belongs here but fixes issue 1013.
+                    // It is not clear what happens with this specific document 'document_with_failed_fonts.pdf'
+                    // I could not find where the same logic is applied in pdfbox.
+                    scanner.Seek(obj.Item2);
+                }
+
                 scanner.MoveNext();
 
                 var token = scanner.CurrentToken;