Allow lenient parsing in DictionaryTokenizer and fix #791

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

@@ -1,5 +1,6 @@
 namespace UglyToad.PdfPig.Tests.Integration
 {
+    using PdfPig.Core;
     using PdfPig.Tokens;
     using Xunit;
 
@@ -33,5 +34,53 @@
                 Assert.Equal("Another Property Value", ((StringToken)valueToken2).Data);
             }
         }
+
+        [Fact]
+        public void CanReadInvalidDocumentInformation()
+        {
+            var path = IntegrationHelpers.GetSpecificTestDocumentPath("invalid-pdf-structure-pdfminer-entire-doc.pdf");
+
+            /*
+                <<
+                /Producer (pdfTeX-1.40.21)
+                 Collaborative Neural Rendering Using Anime Character Sheets /Author()/Title()/Subject()/Creator(LaTeX with hyperref)/Keywords()
+                /CreationDate (D:20230418010134Z)
+                /ModDate (D:20230418010134Z)
+                /Trapped /False
+                /PTEX.Fullbanner (This is pdfTeX, Version 3.14159265-2.6-1.40.21 (TeX Live 2020) kpathsea version 6.3.2)
+                >>
+             */
+
+            // Lenient Parsing On -> can process
+            using (var document = PdfDocument.Open(path))
+            {
+                var information = document.Information;
+
+                Assert.Equal("LaTeX with hyperref", information.Creator);
+                Assert.Equal("", information.Keywords);
+                Assert.Equal("pdfTeX-1.40.21", information.Producer);
+                Assert.Equal("", information.Subject);
+                Assert.Equal("", information.Title);
+                Assert.Equal("", information.Author);
+                Assert.Equal("D:20230418010134Z", information.CreationDate);
+                Assert.Equal("D:20230418010134Z", information.ModifiedDate);
+
+                var infoDictionary = information.DocumentInformationDictionary;
+
+                var nameToken = NameToken.Create("Trapped");
+                Assert.True(infoDictionary.TryGet(nameToken, out var valueToken));
+                Assert.IsType<NameToken>(valueToken);
+                Assert.Equal("False", ((NameToken)valueToken).Data);
+
+                nameToken = NameToken.Create("PTEX.Fullbanner");
+                Assert.True(infoDictionary.TryGet(nameToken, out var valueToken2));
+                Assert.IsType<StringToken>(valueToken2);
+                Assert.Equal("This is pdfTeX, Version 3.14159265-2.6-1.40.21 (TeX Live 2020) kpathsea version 6.3.2", ((StringToken)valueToken2).Data);
+            }
+
+            // Lenient Parsing Off -> throws
+            var ex = Assert.Throws<PdfDocumentFormatException>(() => PdfDocument.Open(path, ParsingOptions.LenientParsingOff));
+            Assert.Equal("Expected name as dictionary key, instead got: Collaborative", ex.Message);
+        }
     }
 }

src/UglyToad.PdfPig.Tests/Tokenization/Scanner/PdfTokenScannerTests.cs

@@ -499,7 +499,7 @@ endobj";
             var input = StringBytesTestConverter.Convert(s, false);
 
             return new PdfTokenScanner(input.Bytes, locationProvider ?? new TestObjectLocationProvider(),
-                new TestFilterProvider(), NoOpEncryptionHandler.Instance);
+                new TestFilterProvider(), NoOpEncryptionHandler.Instance, ParsingOptions.LenientParsingOff);
         }
 
         private static IReadOnlyList<ObjectToken> ReadToEnd(PdfTokenScanner scanner)

src/UglyToad.PdfPig.Tokenization/DictionaryTokenizer.cs

@@ -9,6 +9,7 @@
     {
         private readonly bool usePdfDocEncoding;
         private readonly IReadOnlyList<NameToken> requiredKeys;
+        private readonly bool useLenientParsing;
 
         public bool ReadsNextByte { get; } = false;
 
@@ -22,10 +23,12 @@
         /// Can be provided to recover from errors with missing dictionary end symbols if the
         /// set of keys expected in the dictionary are known.
         /// </param>
-        public DictionaryTokenizer(bool usePdfDocEncoding, IReadOnlyList<NameToken> requiredKeys = null)
+        /// <param name="useLenientParsing">Whether to use lenient parsing.</param>
+        public DictionaryTokenizer(bool usePdfDocEncoding, IReadOnlyList<NameToken> requiredKeys = null, bool useLenientParsing = false)
         {
             this.usePdfDocEncoding = usePdfDocEncoding;
             this.requiredKeys = requiredKeys;
+            this.useLenientParsing = useLenientParsing;
         }
 
         public bool TryTokenize(byte currentByte, IInputBytes inputBytes, out IToken token)
@@ -80,7 +83,7 @@
                 return false;
             }
 
-            var coreScanner = new CoreTokenScanner(inputBytes, usePdfDocEncoding, ScannerScope.Dictionary);
+            var coreScanner = new CoreTokenScanner(inputBytes, usePdfDocEncoding, ScannerScope.Dictionary, useLenientParsing: useLenientParsing);
 
             var tokens = new List<IToken>();
 
@@ -96,7 +99,7 @@
                 // Has enough key/values for each required key
                 if (useRequiredKeys && tokens.Count >= requiredKeys.Count * 2)
                 {
-                    var proposedDictionary = ConvertToDictionary(tokens);
+                    var proposedDictionary = ConvertToDictionary(tokens, useLenientParsing);
 
                     var isAcceptable = true;
                     foreach (var key in requiredKeys)
@@ -118,15 +121,14 @@
                 }
             }
 
-            var dictionary = ConvertToDictionary(tokens);
+            var dictionary = ConvertToDictionary(tokens, useLenientParsing);
 
             token = new DictionaryToken(dictionary);
 
             return true;
-
         }
 
-        private static Dictionary<NameToken, IToken> ConvertToDictionary(List<IToken> tokens)
+        private static Dictionary<NameToken, IToken> ConvertToDictionary(List<IToken> tokens, bool useLenientParsing)
         {
             var result = new Dictionary<NameToken, IToken>();
 
@@ -143,6 +145,13 @@
                         continue;
                     }
 
+                    if (useLenientParsing)
+                    {
+                        // TODO - Log warning
+                        System.Diagnostics.Debug.WriteLine($"Expected name as dictionary key, instead got: " + token);
+                        continue;
+                    }
+
                     throw new PdfDocumentFormatException($"Expected name as dictionary key, instead got: " + token);
                 }
 

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

@@ -27,7 +27,8 @@
         private readonly IInputBytes inputBytes;
         private readonly bool usePdfDocEncoding;
         private readonly List<(byte firstByte, ITokenizer tokenizer)> customTokenizers = new List<(byte, ITokenizer)>();
-        
+        private readonly bool useLenientParsing;
+
         /// <summary>
         /// The offset in the input data at which the <see cref="CurrentToken"/> starts.
         /// </summary>
@@ -52,15 +53,17 @@
             IInputBytes inputBytes,
             bool usePdfDocEncoding,
             ScannerScope scope = ScannerScope.None,
-            IReadOnlyDictionary<NameToken, IReadOnlyList<NameToken>> namedDictionaryRequiredKeys = null)
+            IReadOnlyDictionary<NameToken, IReadOnlyList<NameToken>> namedDictionaryRequiredKeys = null,
+            bool useLenientParsing = false)
         {
             this.inputBytes = inputBytes ?? throw new ArgumentNullException(nameof(inputBytes));
             this.usePdfDocEncoding = usePdfDocEncoding;
             this.stringTokenizer = new StringTokenizer(usePdfDocEncoding);
             this.arrayTokenizer = new ArrayTokenizer(usePdfDocEncoding);
-            this.dictionaryTokenizer = new DictionaryTokenizer(usePdfDocEncoding);
+            this.dictionaryTokenizer = new DictionaryTokenizer(usePdfDocEncoding, useLenientParsing: useLenientParsing);
             this.scope = scope;
             this.namedDictionaryRequiredKeys = namedDictionaryRequiredKeys;
+            this.useLenientParsing = useLenientParsing;
         }
 
         /// <inheritdoc />
@@ -140,7 +143,7 @@
                                     && CurrentToken is NameToken name
                                     && namedDictionaryRequiredKeys.TryGetValue(name, out var requiredKeys))
                                 {
-                                    tokenizer = new DictionaryTokenizer(usePdfDocEncoding, requiredKeys);
+                                    tokenizer = new DictionaryTokenizer(usePdfDocEncoding, requiredKeys, useLenientParsing);
                                 }
                             }
                             else

src/UglyToad.PdfPig/Parser/PageContentParser.cs

@@ -28,7 +28,7 @@
             IInputBytes inputBytes,
             ILog log)
         {
-            var scanner = new CoreTokenScanner(inputBytes, false);
+            var scanner = new CoreTokenScanner(inputBytes, false, useLenientParsing: useLenientParsing);
 
             var precedingTokens = new List<IToken>();
             var graphicsStateOperations = new List<IGraphicsStateOperation>();

src/UglyToad.PdfPig/Parser/PdfDocumentFactory.cs

@@ -75,7 +75,7 @@
                 SkipMissingFonts = false
             };
 
-            var tokenScanner = new CoreTokenScanner(inputBytes, true);
+            var tokenScanner = new CoreTokenScanner(inputBytes, true, useLenientParsing: options.UseLenientParsing);
 
             var passwords = new List<string>();
 
@@ -115,7 +115,7 @@
             // We're ok with this since our intent is to lazily load the cross reference table.
             // ReSharper disable once AccessToModifiedClosure
             var locationProvider = new ObjectLocationProvider(() => crossReferenceTable, inputBytes);
-            var pdfScanner = new PdfTokenScanner(inputBytes, locationProvider, filterProvider, NoOpEncryptionHandler.Instance);
+            var pdfScanner = new PdfTokenScanner(inputBytes, locationProvider, filterProvider, NoOpEncryptionHandler.Instance, parsingOptions);
 
             var crossReferenceStreamParser = new CrossReferenceStreamParser(filterProvider);
             var crossReferenceParser = new CrossReferenceParser(parsingOptions.Logger, xrefValidator, crossReferenceStreamParser);

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

@@ -25,6 +25,7 @@
         private readonly IObjectLocationProvider objectLocationProvider;
         private readonly ILookupFilterProvider filterProvider;
         private readonly CoreTokenScanner coreTokenScanner;
+        private readonly ParsingOptions parsingOptions;
 
         private IEncryptionHandler encryptionHandler;
         private bool isDisposed;
@@ -52,13 +53,14 @@
         public long Length => coreTokenScanner.Length;
 
         public PdfTokenScanner(IInputBytes inputBytes, IObjectLocationProvider objectLocationProvider, ILookupFilterProvider filterProvider,
-            IEncryptionHandler encryptionHandler)
+            IEncryptionHandler encryptionHandler, ParsingOptions parsingOptions)
         {
             this.inputBytes = inputBytes;
             this.objectLocationProvider = objectLocationProvider;
             this.filterProvider = filterProvider;
             this.encryptionHandler = encryptionHandler;
-            coreTokenScanner = new CoreTokenScanner(inputBytes, true);
+            this.parsingOptions = parsingOptions;
+            coreTokenScanner = new CoreTokenScanner(inputBytes, true, useLenientParsing: parsingOptions.UseLenientParsing);
         }
 
         public void UpdateEncryptionHandler(IEncryptionHandler newHandler)
@@ -815,7 +817,7 @@
             // Read the N integers
             var bytes = new ByteArrayInputBytes(stream.Decode(filterProvider, this));
 
-            var scanner = new CoreTokenScanner(bytes, true);
+            var scanner = new CoreTokenScanner(bytes, true, useLenientParsing: parsingOptions.UseLenientParsing);
 
             var objects = new List<Tuple<long, long>>();