gistfile1.cs

using System.Collections.Generic;
using System.Linq;
using HtmlAgilityPack;

namespace TeamVision.Helpers
{
	/*
	 *	remove blacklisted
	 *	sanitize whitelisted sttributes
	 *	inore textnodes
	 *	unwrap others
	 *  map tags
	 *	normalize nested hierarchies
	 *	normalize headings
	 *	wrap all blockless textnodes to paragraphs
	 *	
	 *	todo: test
	 * 
	 *	<h1>Ahoj</h1>
	 *	<h5>do h2</h5>
	 *	<script>alert('xxs')</script>
	 *	<p><div>ho<b>v</b>no</div></p>
	 *	<strong>na b</strong><em>na i</em>
	 * 	<cite>cituju</cite>
	 *	<a style='display: none' href=google.com>foo</a>
	 *	<scrip>alert('xxs')</scrip>
	 *	<!--fuck-->	 
	 * 	
	 *	<h1>Ahoj</h1>
	 *	<h2>do h2</h2>
	 *	<p>ho<b>v</b>no</p>
	 *	<b>na b</b><i>na i</i>
	 *	cituju
	 *	<a href="google.com">foo</a>
	 *	alert('xxs')
	 * 
	 */
	public static class HtmlSanitizer
	{
		static readonly IDictionary<string, string[]> whitelist = new Dictionary<string, string[]>  
		{ 
			{ "h1", null },	
			{ "h2", null },	
			{ "h3", null },	
			{ "h4", null },	
			{ "h5", null },	
			{ "h6", null },	
			{ "a", new[] { "href" } },
			{ "img", new[] { "src", "title" } },
			{ "strong", null },
			{ "b", null },
			{ "em", null },
			{ "i", null },
			{ "p", null },
			{ "blockquote", null },
			{ "ul", null },
			{ "ol", null },
			{ "li", null },
			{ "div", null },
			{ "sub", null }, 
			{ "sup", null }, 
			{ "table", null },
			{ "tr", null },
			{ "td", null },
			{ "th", null }
		};

		static readonly string[] blacklist = new string[] { 
			"script" 
		};

		static readonly IDictionary<string, string> map = new Dictionary<string, string>
		{
			{ "strong", "b"},
			{ "em", "i"},
			{ "div", "p"},
			{ "h4", "h3"},
			{ "h5", "h3"},
			{ "h6", "h3"} 
		};

		static readonly string[] blocks = new string[] { 
			"h1",
			"h2",
			"h3",
			"h4",
			"h5",
			"h6",
			"p",
			"blockquote",
			"li",
			"td",
			"th"
		};

		public static string Sanitize(this string str)
		{
			var doc = new HtmlDocument();
			doc.OptionFixNestedTags = true;
			doc.LoadHtml(str);
			Sanitize(doc);
			return doc.DocumentNode.WriteTo().Trim();
		}

		static void Sanitize(HtmlDocument doc)
		{
			ProcessListed(doc);
			UnwrapDoubleNested(doc);
			WrapBlocklessTextNodes(doc);
		}

		static void ProcessListed(HtmlDocument doc)
		{
			foreach (var node in GetAll(doc))
			{
				switch (node.NodeType)
				{
					case HtmlNodeType.Element:
						if (whitelist.ContainsKey(node.Name))
						{
							node.SanitizeAttributes();
							if (map.ContainsKey(node.Name))
								node.ChangeTag(map[node.Name]);
						}
						else if (blacklist.Contains(node.Name))
							node.Remove();
						else
							node.Unwrap();
						break;
					case HtmlNodeType.Text:
						continue;
					default:
						node.Remove();
						break;
				}
			}
		}

		static void UnwrapDoubleNested(HtmlDocument doc)
		{
			while (true)
			{
				var nested = GetAll(doc).FirstOrDefault(n =>
					n.ParentNode != null &&
					n.ParentNode.Name == n.Name
				);
				if (nested == null)
					break;
				nested.Unwrap();
			};
		}

		static void WrapBlocklessTextNodes(HtmlDocument doc)
		{
			var nodes = GetAll(doc).Where(n =>
				n.NodeType == HtmlNodeType.Text &&
				!n.HasBlockParent()
			);
			foreach (var node in nodes)
			{
				var p = node.OwnerDocument.CreateElement("p");
				node.ParentNode.ReplaceChild(p, node);
				p.AppendChild(node);
			}
		}

		static IEnumerable<HtmlNode> GetAll(HtmlDocument doc)
		{
			return doc.DocumentNode.ChildNodes.Flatten(ch => ch.ChildNodes).ToArray();
		}

		public static void SanitizeAttributes(this HtmlNode node)
		{
			if (node.HasAttributes)
			{
				var allowed = whitelist[node.Name];
				for (var i = node.Attributes.Count - 1; i >= 0; i--)
				{
					var attribute = node.Attributes[i];
					if (allowed == null || !allowed.Contains(attribute.Name))
						node.Attributes.Remove(attribute);
				}
			}
		}

		public static void ChangeTag(this HtmlNode node, string tagName)
		{
			// clone because this doesnt work: node.ParentNode.InsertBefore(node.FirstChild, node)
			// System.InvalidProgramException: Unexpected error.				
			var clone = node.OwnerDocument.CreateElement(tagName);
			foreach (var child in node.ChildNodes)
				clone.AppendChild(child);
			node.ParentNode.ReplaceChild(clone, node);
		}

		public static void Unwrap(this HtmlNode node)
		{
			// clone because this doesnt work: node.ParentNode.InsertBefore(node.FirstChild, node)
			// System.InvalidProgramException: Unexpected error.				
			var clone = node.Clone();
			foreach (var child in clone.ChildNodes)
				node.ParentNode.InsertBefore(child, node);
			node.Remove();
		}

		public static bool HasBlockParent(this HtmlNode node)
		{
			var parent = node.ParentNode;
			while (parent != null) {
				if (blocks.Contains(parent.Name))
					return true;
				parent = parent.ParentNode;
			}
			return false;
		}

	}
}