First draft for incremental saving

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Advanced/PdfCatalog.cs

@@ -68,7 +68,7 @@ public PdfPages Pages
                 if (_pages == null)
                 {
                     _pages = (PdfPages?)Elements.GetValue(Keys.Pages, VCF.CreateIndirect) ?? NRT.ThrowOnNull<PdfPages>();
-                    if (Owner.IsImported)
+                    if (Owner.IsImported && Owner._openMode != PdfDocumentOpenMode.Append)
                         _pages.FlattenPageTree();
                 }
                 return _pages;

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Advanced/PdfContent.cs

@@ -38,7 +38,7 @@ public PdfContent(PdfDictionary dict) // HACK PdfContent
             : base(dict)
         {
             // A PdfContent dictionary is always unfiltered.
-            Decode();
+            Owner.IrefTable.IgnoreModify(Decode);   // decode modifies the object, ignore that
         }
 
         /// <summary>

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Advanced/PdfContents.cs

@@ -49,7 +49,8 @@ public PdfContent AppendContent()
         {
             Debug.Assert(Owner != null);
 
-            SetModified();
+            if (Owner._openMode != PdfDocumentOpenMode.Append)
+                SetModified();
             PdfContent content = new PdfContent(Owner);
             Owner.IrefTable.Add(content);
             Debug.Assert(content.Reference != null);
@@ -64,7 +65,8 @@ public PdfContent PrependContent()
         {
             Debug.Assert(Owner != null);
 
-            SetModified();
+            if (Owner._openMode != PdfDocumentOpenMode.Append)
+                SetModified();
             PdfContent content = new PdfContent(Owner);
             Owner.IrefTable.Add(content);
             Debug.Assert(content.Reference != null);

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Advanced/PdfCrossReferenceTable.cs

@@ -24,11 +24,47 @@ public PdfCrossReferenceTable(PdfDocument document)
         public Dictionary<PdfObjectID, PdfReference> ObjectTable = [];
 
         /// <summary>
+        /// Used to collect modified objects for incremental updates
+        /// </summary>
+        public Dictionary<PdfObjectID, PdfReference> ModifiedObjects = [];
+
         /// Gets or sets a value indicating whether this table is under construction.
         /// It is true while reading a PDF file.
         /// </summary>
         internal bool IsUnderConstruction { get; set; }
 
+        internal bool ReadyForModification { get; set; }
+
+        internal void MarkAsModified(PdfReference? pdfReference)
+        {
+            if (pdfReference == null || !ReadyForModification)
+                return;
+
+            if (pdfReference.ObjectID.IsEmpty)
+                throw new ArgumentException("ObjectID must not be empty", nameof(pdfReference.ObjectID));
+
+            ModifiedObjects[pdfReference.ObjectID] = pdfReference;
+        }
+
+        /// <summary>
+        /// Used to temporarily ignore modifications to objects<br></br>
+        /// (i.e. when doing type-transformations that do not change the structure of the document)
+        /// </summary>
+        /// <param name="action"></param>
+        internal void IgnoreModify(Action action)
+        {
+            var prev = ReadyForModification;
+            ReadyForModification = false;
+            try
+            {
+                action();
+            }
+            finally
+            {
+                ReadyForModification = prev;
+            }
+        }
+
         /// <summary>
         /// Adds a cross-reference entry to the table. Used when parsing the trailer.
         /// </summary>
@@ -46,6 +82,9 @@ public void Add(PdfReference iref)
                 throw new InvalidOperationException("Object already in table.");
 
             ObjectTable.Add(iref.ObjectID, iref);
+
+            if (ReadyForModification && _document.IsAppending)
+                ModifiedObjects[iref.ObjectID] = iref;
         }
 
         /// <summary>
@@ -65,6 +104,9 @@ public void Add(PdfObject value)
                 throw new InvalidOperationException("Object already in table.");
 
             ObjectTable.Add(value.ObjectID, value.ReferenceNotNull);
+
+            if (ReadyForModification && _document.IsAppending)
+                ModifiedObjects[value.ObjectID] = value.Reference;
         }
 
         /// <summary>

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Advanced/PdfTrailer.cs

@@ -14,6 +14,12 @@ namespace PdfSharp.Pdf.Advanced
     // Reference: 3.4.4  File Trailer / Page 96
     class PdfTrailer : PdfDictionary
     {
+        /// <summary>
+        /// Gets or sets the position of this trailer in the input-stream<br></br>
+        /// Only meaningful for loaded documents; will be zero for new documents
+        /// </summary>
+        internal SizeType Position { get; set; }
+
         /// <summary>
         /// Initializes a new instance of PdfTrailer.
         /// </summary>
@@ -211,8 +217,9 @@ internal void Finish()
 
             Elements.Remove(Keys.Prev);
 
-            Debug.Assert(_document.IrefTable.IsUnderConstruction == false);
+            Debug.Assert(_document.IrefTable.IsUnderConstruction == false);     // Why ??
             _document.IrefTable.IsUnderConstruction = false;
+            _document.IrefTable.ReadyForModification = true;
         }
 
         /// <summary>

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.Annotations/PdfAnnotation.cs

@@ -83,7 +83,7 @@ public PdfAnnotations Parent
         /// </summary>
         public PdfRectangle Rectangle
         {
-            get => Elements.GetRectangle(Keys.Rect, true);
+            get => Elements.GetRectangle(Keys.Rect);
             set
             {
                 Elements.SetRectangle(Keys.Rect, value);

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.IO/Parser.cs

@@ -1161,11 +1161,13 @@ internal PdfTrailer ReadTrailer()
             // Read position behind 'startxref'.
             _lexer.Position = ReadSize();
 
+            var xrefStart = _lexer.Position;
+
             // Read all trailers.
             PdfTrailer? newerTrailer = null;
             while (true)
             {
-                var trailer = ReadXRefTableAndTrailer(_document.IrefTable);
+                var trailer = ReadXRefTableAndTrailer(_document.IrefTable, xrefStart);
 
                 // Return the first found trailer, which is the one 'startxref' points to.
                 // This is the current trailer, even for incrementally updated files.
@@ -1183,6 +1185,7 @@ internal PdfTrailer ReadTrailer()
 
                 // Continue loading previous trailer and cache this one as the newerTrailer to add its previous trailer.
                 _lexer.Position = prev;
+                xrefStart = prev;
                 newerTrailer = trailer;
             }
             return _document.Trailer;
@@ -1191,7 +1194,7 @@ internal PdfTrailer ReadTrailer()
         /// <summary>
         /// Reads cross-reference table(s) and trailer(s).
         /// </summary>
-        PdfTrailer? ReadXRefTableAndTrailer(PdfCrossReferenceTable xrefTable)
+        PdfTrailer? ReadXRefTableAndTrailer(PdfCrossReferenceTable xrefTable, SizeType xrefStart)
         {
             Debug.Assert(xrefTable != null);
 
@@ -1260,7 +1263,10 @@ internal PdfTrailer ReadTrailer()
                     else if (symbol == Symbol.Trailer)
                     {
                         ReadSymbol(Symbol.BeginDictionary);
-                        var trailer = new PdfTrailer(_document);
+                        var trailer = new PdfTrailer(_document)
+                        {
+                            Position = xrefStart
+                        };
                         ReadDictionary(trailer, false);
                         return trailer;
                     }
@@ -1275,8 +1281,8 @@ internal PdfTrailer ReadTrailer()
                 // Reference: 3.4.7  Cross-Reference Streams / Page 93
                 // TODO: We have not yet tested PDF files larger than 2 GiB because we have none and cannot produce one.
 
-                // The parsed integer is the object ID of the cross-reference stream object.
-                return ReadXRefStream(xrefTable);
+                // The parsed integer is the object ID of the cross-reference stream.
+                return ReadXRefStream(xrefTable, xrefStart);
             }
             return null;
         }
@@ -1330,14 +1336,11 @@ bool CheckXRefTableEntry(SizeType position, int id, int generation, out int idCh
         /// <summary>
         /// Reads cross-reference stream(s).
         /// </summary>
-        PdfTrailer ReadXRefStream(PdfCrossReferenceTable xrefTable)
+        PdfTrailer ReadXRefStream(PdfCrossReferenceTable xrefTable, SizeType xrefStart)
         {
             // Read cross-reference stream.
             //Debug.Assert(_lexer.Symbol == Symbol.Integer);
 
-            // NEEDED???
-            var xrefStart = _lexer.Position - _lexer.Token.Length;
-
             int number = _lexer.TokenToInteger;
             int generation = ReadInteger();
             // According to specs, generation number "shall not" be "other than zero".
@@ -1357,7 +1360,10 @@ PdfTrailer ReadXRefStream(PdfCrossReferenceTable xrefTable)
             ReadSymbol(Symbol.BeginDictionary);
             var objectID = new PdfObjectID(number, generation);
 
-            var xrefStream = new PdfCrossReferenceStream(_document);
+            var xrefStream = new PdfCrossReferenceStream(_document)
+            {
+                Position = xrefStart
+            };
 
             ReadDictionary(xrefStream, false);
             ReadSymbol(Symbol.BeginStream);

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.IO/PdfReader.cs

@@ -2,6 +2,7 @@
 // See the LICENSE file in the solution root for more information.
 
 using Microsoft.Extensions.Logging;
+using Microsoft.VisualBasic;
 using PdfSharp.Internal;
 using PdfSharp.Logging;
 using PdfSharp.Pdf.Advanced;
@@ -345,7 +346,8 @@ PdfDocument OpenFromStream(Stream stream, string? password, PdfDocumentOpenMode
                                 throw new PdfReaderException(PSSR.InvalidPassword);
                         }
                     }
-                    else if (validity == PasswordValidity.UserPassword && openMode == PdfDocumentOpenMode.Modify)
+                    else if (validity == PasswordValidity.UserPassword 
+                        && (openMode == PdfDocumentOpenMode.Modify || openMode == PdfDocumentOpenMode.Append))
                     {
                         if (passwordProvider != null)
                         {
@@ -445,16 +447,19 @@ void FinishReferences()
                     "All references saved in IrefTable should have been created when their referred PdfObject has been accessible.");
 
                 // Get and update object's references.
-                FinishItemReferences(iref.Value, _document, finishedObjects);
+                FinishItemReferences(iref.Value, iref, _document, finishedObjects);
             }
 
+            // why setting it here AND in Trailer.Finish ??
             _document.IrefTable.IsUnderConstruction = false;
 
             // Fix references of trailer values and then objects and irefs are consistent.
             _document.Trailer.Finish();
+
+            Debug.Assert(_document.IrefTable.ModifiedObjects.Count == 0, "There should be no modified objects");
         }
 
-        void FinishItemReferences(PdfItem? pdfItem, PdfDocument document, HashSet<PdfObject> finishedObjects)
+        void FinishItemReferences(PdfItem? pdfItem, PdfReference itemReference, PdfDocument document, HashSet<PdfObject> finishedObjects)
         {
             // Only PdfObjects may contain further PdfReferences.
             if (pdfItem is not PdfObject pdfObject)
@@ -478,10 +483,12 @@ void FinishItemReferences(PdfItem? pdfItem, PdfDocument document, HashSet<PdfObj
             switch (pdfObject)
             {
                 case PdfDictionary childDictionary:
-                    FinishChildReferences(childDictionary, finishedObjects);
+                    FinishChildReferences(childDictionary, childDictionary.Reference ?? itemReference, finishedObjects);
+                    childDictionary.SetModified(false);
                     break;
                 case PdfArray childArray:
-                    FinishChildReferences(childArray, finishedObjects);
+                    FinishChildReferences(childArray, childArray.Reference ?? itemReference, finishedObjects);
+                    childArray.SetModified(false);
                     break;
             }
 #else
@@ -493,8 +500,13 @@ void FinishItemReferences(PdfItem? pdfItem, PdfDocument document, HashSet<PdfObj
 #endif
         }
 
-        void FinishChildReferences(PdfDictionary dictionary, HashSet<PdfObject> finishedObjects)
+        void FinishChildReferences(PdfDictionary dictionary, PdfReference containingReference, HashSet<PdfObject> finishedObjects)
         {
+            if (dictionary.ObjectNumber == 15)
+                GetType();
+            if (dictionary.Reference is null && dictionary.ContainingReference is null)
+                dictionary.ContainingReference = containingReference;
+
             // Dictionary elements are modified inside loop. Avoid "Collection was modified; enumeration operation may not execute" error occuring in net 4.7.2.
             // There is no way to access KeyValuePairs via index natively to use a for loop with.
             // Instead, enumerate Keys and get value via Elements[key], which shall be O(1).
@@ -511,12 +523,15 @@ void FinishChildReferences(PdfDictionary dictionary, HashSet<PdfObject> finished
                 }
 
                 // Get and update item's references.
-                FinishItemReferences(item, _document, finishedObjects);
+                FinishItemReferences(item, containingReference, _document, finishedObjects);
             }
         }
 
-        void FinishChildReferences(PdfArray array, HashSet<PdfObject> finishedObjects)
+        void FinishChildReferences(PdfArray array, PdfReference containingReference, HashSet<PdfObject> finishedObjects)
         {
+            if (array.Reference is null && array.ContainingReference is null)
+                array.ContainingReference = containingReference;
+
             var elements = array.Elements;
             for (var i = 0; i < elements.Count; i++)
             {
@@ -531,7 +546,7 @@ void FinishChildReferences(PdfArray array, HashSet<PdfObject> finishedObjects)
                 }
 
                 // Get and update item's references.
-                FinishItemReferences(item, _document, finishedObjects);
+                FinishItemReferences(item, containingReference, _document, finishedObjects);
             }
         }
 

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.IO/PdfWriter.cs

@@ -524,7 +524,8 @@ public void WriteEof(PdfDocument document, SizeType startxref)
             WriteRaw(startxref.ToString(CultureInfo.InvariantCulture));
             WriteRaw("\n%%EOF\n");
             SizeType fileSize = (SizeType)_stream.Position;
-            if (_layout == PdfWriterLayout.Verbose)
+            // position check required for incremental updates to avoid overwriting the start of the file
+            if (_layout == PdfWriterLayout.Verbose && _commentPosition > 0)
             {
                 TimeSpan duration = DateTime.Now - document._creation;
 

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.IO/enums/PdfDocumentOpenMode.cs

@@ -33,5 +33,10 @@ public enum PdfDocumentOpenMode
         /// and is e.g. useful for browsing information about a collection of PDF documents in a user interface.
         /// </summary>
         InformationOnly,  // TODO: not yet implemented
+
+        /// <summary>
+        /// Comparable to <see cref="Modify"/> but changes are appended to the document when saving
+        /// </summary>
+        Append
     }
 }