Fix bug where FormXObject bbox needs to be normalised

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

@@ -1,5 +1,7 @@
 namespace UglyToad.PdfPig.Tests.Integration
 {
+    using DocumentLayoutAnalysis.PageSegmenter;
+    using DocumentLayoutAnalysis.WordExtractor;
     using UglyToad.PdfPig.Core;
 
     public class XObjectFormTests
@@ -32,5 +34,24 @@ public void CanReadDocumentMOZILLA_3136_0()
                 var page = document.GetPage(1);
             }
         }
+
+        [Fact]
+        public void XObjectFormXClipping()
+        {
+            using (var document = PdfDocument.Open(IntegrationHelpers.GetDocumentPath("ICML03-081"), new ParsingOptions()
+                   {
+                       ClipPaths = true
+                   }))
+            {
+                var page = document.GetPage(4);
+                Assert.True(page.Paths.Count > 3);
+
+                var words = page.GetWords(NearestNeighbourWordExtractor.Instance);
+                var blocks = DocstrumBoundingBoxes.Instance.GetBlocks(words).SelectMany(b => b.TextLines).ToArray();
+
+                Assert.Equal(2, blocks.Count(l => l.Text.Equals("Training Examples per Class")));
+                Assert.Equal(2, blocks.Count(l => l.Text.Equals("Classification Weight")));
+            }
+        }
     }
 }

src/UglyToad.PdfPig/Graphics/BaseStreamProcessor.cs

@@ -576,7 +576,7 @@ protected virtual void ProcessFormXObject(StreamToken formStream, NameToken xObj
             {
                 var points = bboxToken.Data.OfType<NumericToken>().Select(x => x.Double).ToArray();
                 PdfRectangle bbox = new PdfRectangle(points[0], points[1], points[2], points[3]);
-                PdfRectangle transformedBox = startState.CurrentTransformationMatrix.Transform(bbox);
+                PdfRectangle transformedBox = startState.CurrentTransformationMatrix.Transform(bbox).Normalise();
                 ClipToRectangle(transformedBox, FillingRule.EvenOdd); // TODO - Check that Even Odd is valid
             }