From c3cccc3f31b10e780029e6f452fbec5f9e48beb2 Mon Sep 17 00:00:00 2001
From: Tobias Klika <cee@live.at>
Date: Fri, 31 Jan 2014 10:50:44 +0100
Subject: [PATCH] remove unnecessary attributes from XElements

---
 .../NReadability/NReadabilityTranscoder.cs    | 35 +++++++++++++++++++
 ReadSharp.Tests/ReadTests.cs                  | 14 ++++----
 2 files changed, 42 insertions(+), 7 deletions(-)

diff --git a/PortablePorts/NReadability/NReadabilityTranscoder.cs b/PortablePorts/NReadability/NReadabilityTranscoder.cs
index bb58c0a..7f5d53f 100644
--- a/PortablePorts/NReadability/NReadabilityTranscoder.cs
+++ b/PortablePorts/NReadability/NReadabilityTranscoder.cs
@@ -126,6 +126,7 @@ namespace ReadSharp.Ports.NReadability
     private static readonly Regex _PageRegex = new Regex("pag(e|ing|inat)|([^a-z]|^)pag([^a-z]|$)", RegexOptions.IgnoreCase);
     private static readonly Regex _LikelyParagraphDivRegex = new Regex("text|para|parbase|paragraph|figure", RegexOptions.IgnoreCase);
     private static readonly Regex _LikelyImageContainerRegex = new Regex("image|figure|photo|media", RegexOptions.IgnoreCase);
+    private static readonly Regex _ReplaceAttributes = new Regex("^(data-|aria-|name|height|width|item|value|contenteditable|on|target|type)", RegexOptions.IgnoreCase);
 
     #endregion
 
@@ -1303,6 +1304,8 @@ namespace ReadSharp.Ports.NReadability
 
       RemoveElements(elementsToRemove);
 
+      CleanAttributes(articleContentElement);
+
       /* Remove br's that are directly before paragraphs. */
       articleContentElement.SetInnerHtml(_BreakBeforeParagraphRegex.Replace(articleContentElement.GetInnerHtml(), "<p"));
     }
@@ -1571,6 +1574,33 @@ namespace ReadSharp.Ports.NReadability
         }).Traverse(rootElement);
     }
 
+
+    internal void CleanAttributes(XElement rootElement)
+    {
+      new ElementsTraverser(
+        element =>
+        {
+          IEnumerable<XAttribute> attributes = element.Attributes();
+          List<XAttribute> attributesToRemove = new List<XAttribute>();
+
+          if (attributes == null)
+          {
+            return;
+          }
+
+
+          foreach (XAttribute attribute in attributes)
+          {
+            if (_ReplaceAttributes.IsMatch(attribute.Name.ToString()))
+            {
+              attributesToRemove.Add(attribute);
+            }
+          }
+
+          RemoveAttributes(attributesToRemove);
+        }).Traverse(rootElement);
+    }
+
     internal string GetUserStyleClass(string prefix, String enumStr)
     {
       var suffixSB = new StringBuilder();
@@ -1632,6 +1662,11 @@ namespace ReadSharp.Ports.NReadability
       elementsToRemove.ForEach(elementToRemove => elementToRemove.Remove());
     }
 
+    private static void RemoveAttributes(IEnumerable<XAttribute> attributesToRemove)
+    {
+      attributesToRemove.ForEach(attributeToRemove => attributeToRemove.Remove());
+    }
+
     private static void ResolveElementsUrls(XDocument document, string tagName, string attributeName, string url, Func<AttributeTransformationInput, AttributeTransformationResult> attributeValueTransformer)
     {
       if (document == null)
diff --git a/ReadSharp.Tests/ReadTests.cs b/ReadSharp.Tests/ReadTests.cs
index 8bbd176..4204969 100644
--- a/ReadSharp.Tests/ReadTests.cs
+++ b/ReadSharp.Tests/ReadTests.cs
@@ -138,16 +138,16 @@ namespace ReadSharp.Tests
     [Fact]
     public async Task TestCriticalURIs()
     {
-      //Article result = await reader.Read(new Uri("http://wpcentral.com.feedsportal.com/c/33999/f/616880/s/35a02b5e/sc/15/l/0L0Swpcentral0N0Cgameloft0Ediscusses0Etheir0Enew0Egame0Ebrothers0Earms0E30Esons0Ewar0Eceslive/story01.htm"));
-      //Assert.NotEmpty(result.Content);
+      Article result = await reader.Read(new Uri("http://wpcentral.com.feedsportal.com/c/33999/f/616880/s/35a02b5e/sc/15/l/0L0Swpcentral0N0Cgameloft0Ediscusses0Etheir0Enew0Egame0Ebrothers0Earms0E30Esons0Ewar0Eceslive/story01.htm"));
+      Assert.NotEmpty(result.Content);
 
-      //result = await reader.Read(new Uri("http://msdn.microsoft.com/en-us/library/windows/apps/hh464925.aspx"));
-      //Assert.NotEmpty(result.Content);
+      result = await reader.Read(new Uri("http://msdn.microsoft.com/en-us/library/windows/apps/hh464925.aspx"));
+      Assert.NotEmpty(result.Content);
 
-      //result = await reader.Read(new Uri("http://bit.ly/KAh7FJ"));
-      //Assert.NotEmpty(result.Content);
+      result = await reader.Read(new Uri("http://bit.ly/KAh7FJ"));
+      Assert.NotEmpty(result.Content);
 
-      Article result = await reader.Read(new Uri("http://www.nytimes.com/2014/01/31/world/europe/ukraine-unrest.html?hp&_r=0"));
+      result = await reader.Read(new Uri("http://www.nytimes.com/2014/01/31/world/europe/ukraine-unrest.html?hp&_r=0"));
       Assert.True(result.Images != null && result.Images.Count > 0 && result.Images.SingleOrDefault(item => item.Uri.OriginalString == "http://static01.nyt.com/images/2014/01/31/world/31ukraine-cnd01/31ukraine-cnd01-articleLarge.jpg") != null);
     }