Merge pull request #447 from jltrem/jltrem/pdf-image-consolidation

Add image consolidation

PdfSharpCore.Test/Merge.cs

@@ -1,25 +1,74 @@
-using System.IO;
+using System;
+using System.Collections.Generic;
+using System.IO;
+using System.Linq;
+using PdfSharpCore.Drawing;
+using PdfSharpCore.Drawing.Layout;
 using PdfSharpCore.Pdf;
 using PdfSharpCore.Pdf.IO;
 using PdfSharpCore.Test.Helpers;
 using Xunit;
+using Xunit.Abstractions;
 
 namespace PdfSharpCore.Test
 {
     public class Merge
     {
+        private readonly ITestOutputHelper _output;
+
+        public Merge(ITestOutputHelper output)
+        {
+            _output = output;
+        }
+
         [Fact]
         public void CanMerge2Documents()
         {
             var pdf1Path =  PathHelper.GetInstance().GetAssetPath("FamilyTree.pdf");
             var pdf2Path = PathHelper.GetInstance().GetAssetPath("test.pdf");
 
+            var outputDocument = MergeDocuments(new[] { pdf1Path, pdf2Path });
+
+            var outFilePath = CreateOutFilePath("merge.pdf");
+            outputDocument.Save(outFilePath);
+        }
+
+        [Fact]
+        public void CanConsolidateImageDataInDocument()
+        {
+            var doc1 = CreateTestDocumentWithImage("lenna.png");
+            var doc2 = CreateTestDocumentWithImage("frog-and-toad.jpg");
+
+            var pdf1Path = CreateOutFilePath("image-doc1.pdf");
+            doc1.Save(pdf1Path);
+
+            var pdf2Path = CreateOutFilePath("image-doc2.pdf");
+            doc2.Save(pdf2Path);
+
+            var pdfPathsForMerge = Enumerable.Range(1, 50).SelectMany(_ => new[] { pdf1Path, pdf2Path });
+            var outputDocument = MergeDocuments(pdfPathsForMerge);
+
+            var mergedFilePath = CreateOutFilePath("images-merged.pdf");
+            outputDocument.Save(mergedFilePath);
+
+            outputDocument.ConsolidateImages();
+            var consolidatedFilePath = CreateOutFilePath("images-merged-consolidated.pdf");
+            outputDocument.Save(consolidatedFilePath);
+
+            long mergedLength = new FileInfo(mergedFilePath).Length;
+            long consolidatedLength = new FileInfo(consolidatedFilePath).Length;
+            Assert.True(consolidatedLength < mergedLength / 4);
+        }
+
+        private static PdfDocument MergeDocuments(IEnumerable<string> pdfPaths)
+        {
             var outputDocument = new PdfDocument();
 
-            foreach (var pdfPath in new[] { pdf1Path, pdf2Path })
+            foreach (var pdfPath in pdfPaths)
             {
                 using var fs = File.OpenRead(pdfPath);
                 var inputDocument = Pdf.IO.PdfReader.Open(fs, PdfDocumentOpenMode.Import);
+
                 var count = inputDocument.PageCount;
                 for (var idx = 0; idx < count; idx++)
                 {
@@ -28,14 +77,34 @@ public void CanMerge2Documents()
                 }
             }
 
-            var outFilePath = Path.Combine(PathHelper.GetInstance().RootDir, "Out", "merge.pdf");
+            return outputDocument;
+        }
+
+        private static string CreateOutFilePath(string filename)
+        {
+            var outFilePath = Path.Combine(PathHelper.GetInstance().RootDir, "Out", filename);
             var dir = Path.GetDirectoryName(outFilePath);
             if (!Directory.Exists(dir))
             {
                 Directory.CreateDirectory(dir);
             }
 
-            outputDocument.Save(outFilePath);
+            return outFilePath;
+        }
+
+        private static PdfDocument CreateTestDocumentWithImage(string imageFilename)
+        {
+            var document = new PdfDocument();
+
+            var pageNewRenderer = document.AddPage();
+            var renderer = XGraphics.FromPdfPage(pageNewRenderer);
+            var textFormatter = new XTextFormatter(renderer);
+
+            var layout = new XRect(12, 12, 400, 50);
+            textFormatter.DrawString(imageFilename, new XFont("Arial", 12), XBrushes.Black, layout);
+            renderer.DrawImage(XImage.FromFile(PathHelper.GetInstance().GetAssetPath(imageFilename)), new XPoint(12, 100));
+
+            return document;
         }
     }
 }

PdfSharpCore.Test/PdfSharpCore.Test.csproj

@@ -38,6 +38,9 @@
     <None Update="Assets\**\*.png">
       <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
     </None>
+    <None Update="Assets\frog-and-toad.jpg">
+      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
+    </None>
   </ItemGroup>
 
 </Project>

PdfSharpCore/Pdf/PdfDocument.cs

@@ -28,9 +28,12 @@
 #endregion
 
 using System;
+using System.Collections.Generic;
 using System.Diagnostics;
 using System.IO;
 using System.Linq;
+using System.Security.Cryptography;
+using System.Text;
 using PdfSharpCore.Pdf.Advanced;
 using PdfSharpCore.Pdf.Internal;
 using PdfSharpCore.Pdf.IO;
@@ -809,6 +812,83 @@ public void MakeAcroFormsReadOnly()
             }
         }
 
+        public void ConsolidateImages()
+        {
+            var images = ImageInfo.FindAll(this);
+
+            var mapHashcodeToMd5 = new Dictionary<int, string>();
+            var mapMd5ToPdfItem = new Dictionary<string, PdfItem>();
+
+            // Calculate MD5 for each image XObject and build lookups for all images.
+            foreach (ImageInfo img in images)
+            {
+                mapHashcodeToMd5[img.XObject.GetHashCode()] = img.XObjectMD5;
+                mapMd5ToPdfItem[img.XObjectMD5] = img.Item.Value;
+            }
+
+            // Set the PdfItem for each image to the one chosen for the MD5.
+            foreach (ImageInfo img in images)
+            {
+                string md5 = mapHashcodeToMd5[img.XObject.GetHashCode()];
+                img.XObjects.Elements[img.Item.Key] = mapMd5ToPdfItem[md5];
+            }
+        }
+
+        internal class ImageInfo
+        {
+            public PdfDictionary XObjects { get; }
+            public KeyValuePair<string, PdfItem> Item  { get; }
+            public PdfDictionary XObject { get; }
+            public string XObjectMD5 { get; }
+
+            private static readonly MD5 Hasher = MD5.Create();
+
+            public ImageInfo(PdfDictionary xObjects, KeyValuePair<string, PdfItem> item, PdfDictionary xObject)
+            {
+                XObjects = xObjects;
+                Item = item;
+                XObject = xObject;
+                XObjectMD5 = ComputeMD5(xObject.Stream.Value);
+            }
+
+            /// <summary>
+            /// Get info for each image in the document.
+            /// </summary>
+            internal static List<ImageInfo> FindAll(PdfDocument doc) =>
+                doc.Pages.Cast<PdfPage>()
+                    .Select(page => page.Elements.GetDictionary("/Resources"))
+                    .Select(resources => resources?.Elements?.GetDictionary("/XObject"))
+                    .Where(xObjects => xObjects?.Elements != null)
+                    .SelectMany(xObjects =>
+                        from item in xObjects.Elements
+                        let xObject = (item.Value as PdfReference)?.Value as PdfDictionary
+                        where xObject?.Elements?.GetString("/Subtype") == "/Image"
+                        select new ImageInfo(xObjects, item, xObject)
+                    )
+                    .ToList();
+
+            /// <summary>
+            /// Compute and return the MD5 hash of the input data.
+            /// </summary>
+            internal static string ComputeMD5(byte[] input)
+            {
+                byte[] hashBytes;
+                lock (Hasher)
+                {
+                    hashBytes = Hasher.ComputeHash(input);
+                    Hasher.Initialize();
+                }
+
+                var sb = new StringBuilder();
+                foreach (var x in hashBytes)
+                {
+                    sb.Append(x.ToString("x2"));
+                }
+
+                return sb.ToString();
+            }
+        }
+
         /// <summary>
         /// Gets the security handler.
         /// </summary>