private static void AddNodes(HtmlElement parent, string html, bool removeInstructions)
{
int len = html.Length;
int i = -1;
string text = "";
while (i < len - 1)
{
i++;
if (i >= html.Length)
break;
string s = html.Substring(i, 1);
if (s == "<")
{
int start = i + 1;
int stop = html.IndexOf(">", i, StringComparison.Ordinal);
if (stop < i)
continue;
if (text.Trim() != "")
parent.Add(new HtmlText(text));
text = "";
string fragment = html.Substring(start, stop - start);
//html comment
if (fragment.StartsWith("!--") && fragment.EndsWith("--"))
{
parent.Add(new HtmlComment(fragment.Substring(3).Substring(0, fragment.Length - 5)));
i = stop;
}
//html instruction
else if (fragment.StartsWith("!"))
{
if (!removeInstructions)
parent.Add(new HtmlInstruction(fragment.Substring(1).Substring(0, fragment.Length - 1)));
i = stop;
}
else if (fragment.StartsWith("/"))
{
//remove invalid closing tags
i = stop;
}
else
{
HtmlElement node = CreateNode(html, start, ref stop);
parent.Add(node);
i = stop + (node.IsClosed ? 0 : (node.Name.Length + 2));
}
continue;
}
text += s;
}
if (text.Trim() != string.Empty)
parent.Add(new HtmlText(text));
}
public static string StripHtml(HtmlElement element)
{
StringBuilder builder = new StringBuilder();
StripHtml(builder, element);
string result = builder.ToString();
return Regex.Replace(result, @" |\s+", " ");
}
private static bool HasContent(HtmlElement element)
{
if (element.Name == "img" || element.Name == "iframe" || element.Name == "frame")
return true;
// пустой element без контейнеров и текста
if (element.ChildNodes.Count == 0)
return false;
// проверить вложенные элементы и удалить пустые
for (int i = element.ChildNodes.Count - 1; i >= 0; i--)
{
if (element.ChildNodes[i] is HtmlElement)
{
HtmlElement child = element.ChildNodes[i] as HtmlElement;
bool hasContent = HasContent(child);
if (!hasContent)
{
element.ChildNodes.RemoveAt(i);
}
}
else if (element.ChildNodes[i] is HtmlComment
|| element.ChildNodes[i] is HtmlInstruction)
{
element.ChildNodes.RemoveAt(i);
}
}
// если нету вложенныех элементов (элементов контейнеров и элементов текстов)
if (element.ChildNodes.Count == 0)
return false;
// проверить тексты (если есть хоть один контейнер, то ищём контент в нём)
if (!element.ChildNodes.Any(p => p is HtmlElement))
{
bool hasText = HasText(element);
if (!hasText)
{
return false;
}
}
return true;
}
public static void ReplaceMedia(HtmlElement element)
{
for (int i = 0; i < element.ChildNodes.Count; i++)
{
if (element.ChildNodes[i] is HtmlElement)
{
HtmlElement innerElement = element.ChildNodes[i] as HtmlElement;
if (innerElement.Name == "img")
{
HtmlAttribute srcAttribute = innerElement.Attributes.FirstOrDefault(p => p.Name == "src");
Uri address;
if (srcAttribute != null &&
!string.IsNullOrEmpty(srcAttribute.Value) &&
Uri.TryCreate(srcAttribute.Value, UriKind.RelativeOrAbsolute, out address))
{
if (!address.PathAndQuery.StartsWith("/scripts/ckeditor/plugins/smiley/images/"))
{
innerElement.Add("data-original", srcAttribute.Value);
innerElement.Add("src", "/images/gimp_5834.png");
innerElement.Add("style", "");
innerElement.Add("width", "");
innerElement.Add("height", "");
innerElement.Add("class", "art-img-thumb");
}
}
}
else if (innerElement.Name == "iframe" || innerElement.Name == "frame")
{
element.RemoveChildAt(i);
HtmlElement newImage = new HtmlElement("img", true);
newImage.Add("src", "/images/youtube-icon.png");
newImage.Add("class", "art-img-thumb-yt");
element.Insert(i, newImage);
}
else
{
ReplaceMedia(innerElement);
}
}
}
}
//поиск изображений
public static List<HtmlElement> FindElementsOfType(HtmlElement root, string type)
{
List<HtmlElement> images = new List<HtmlElement>();
for (int i = 0; i < root.ChildNodes.Count; i++)
{
if (root.ChildNodes[i] is HtmlElement)
{
HtmlElement child = root.ChildNodes[i] as HtmlElement;
if (child.Name == type)
{
images.Add(child);
}
images.AddRange(FindElementsOfType(child, type));
}
}
return images;
}
private static HtmlElement CreateNode(string htmlPart)
{
string html = htmlPart.Trim();
bool isClosed = html.EndsWith("/");
int spaceIndex = html.IndexOf(" ", StringComparison.Ordinal);
if (spaceIndex < 0)
{
if(isClosed)
html = html.Substring(0, html.Length - 1);
return new HtmlElement(html, !CanHaveContent(html));
}
string tagName = html.Substring(0, spaceIndex);
html = html.Substring(spaceIndex + 1);
if (isClosed)
html = html.Substring(0, html.Length - 1);
if (!isClosed && !CanHaveContent(tagName.ToLower()))
isClosed = true;
HtmlElement element = new HtmlElement(tagName, isClosed);
int len = html.Length;
string attrName = string.Empty;
string attrVal = string.Empty;
bool isName = true;
bool isValue = false;
bool hasQuotes = false;
for (var i = 0; i < len; i++)
{
string s = html.Substring(i, 1);
if (s == "=" && !isValue)
{
isName = false;
isValue = true;
string nextChar = html.Substring(i + 1, 1);
hasQuotes = (nextChar == "\"" || nextChar == "'");
}
else if (s == " " && isName)
{
//add attribute that requires no value
element.Add(attrName, attrName);
//reset attribute name
attrName = string.Empty;
attrVal = string.Empty;
}
else if (s == " " && attrVal.Length > 0)
{
if (!hasQuotes || (attrVal[0] == attrVal[attrVal.Length - 1]))
{
isValue = false;
isName = true;
string value = FixAttributeValue(attrVal);
element.Add(attrName, value);
attrName = string.Empty;
attrVal = string.Empty;
}
else if (isValue)
attrVal += s;
}
else if (isName)
{
attrName += s;
}
else if (isValue)
attrVal += s;
}
if (isName && attrName != string.Empty && attrVal == string.Empty)
attrVal = attrName;
if (attrName != string.Empty && attrVal != string.Empty)
{
string value = FixAttributeValue(attrVal);
element.Add(attrName, value);
}
return element;
}
public HtmlImageInfo(HtmlElement element)
{
Element = element;
_src = element.Attributes.FirstOrDefault(p => p.Name == "src");
}
public void Sanitize(HtmlElement element)
{
IEnumerable<HtmlAttribute> attributesToRemove =
from attribute in element.Attributes
where IsUnSafe(attribute)
select attribute;
element.RemoveAll(attributesToRemove);
for (var i = element.ChildNodes.Count - 1; i >= 0; i--)
{
HtmlNode childNode = element.ChildNodes[i];
if (IsUnSafe(childNode)
|| childNode is HtmlInstruction
|| childNode is HtmlComment)
{
element.ChildNodes.RemoveAt(i);
}
}
foreach (HtmlNode node in element.ChildNodes)
{
if (node is HtmlElement)
Sanitize(node as HtmlElement);
}
}
public static void RemoveEmptyTags(HtmlElement document)
{
bool hasContent = HasContent(document);
}
private static void StripHtml(StringBuilder builder, HtmlElement element)
{
foreach (HtmlNode node in element.ChildNodes)
{
if (node is HtmlText)
{
builder.Append(node.ToString());
}
else if (node is HtmlElement)
{
HtmlElement child = node as HtmlElement;
if(SKIP_TAGS.Contains(child.Name))
{
AddWhitespace(builder);
}
else if (NEW_LINE_TAGS.Contains(child.Name))
{
AddWhitespace(builder);
StripHtml(builder, child);
}
}
}
}
private static bool HasText(HtmlElement element)
{
StringBuilder content = new StringBuilder();
for (int j = 0; j < element.ChildNodes.Count; j++)
{
if (element.ChildNodes[j] is HtmlText)
{
string nodeText = element.ChildNodes[j].ToString().ToLower();
content.Append(nodeText);
}
}
string whiteSpacesExpression = @" |\s+|\t|\n|\r|" + Environment.NewLine;
string pureText = Regex.Replace(content.ToString(), whiteSpacesExpression, string.Empty);
return pureText != string.Empty;
}
public static List<HtmlImageInfo> FindImages(HtmlElement root)
{
return FindElementsOfType(root, "img")
.Select(p => new HtmlImageInfo(p))
.ToList();
}
