public static string ExtractTextFromHtml(string html, bool trimWhiteSpace)
{
string str = new Regex("<!*[^<>]*>", RegexOptions.IgnorePatternWhitespace | RegexOptions.Multiline | RegexOptions.IgnoreCase).Replace(html, string.Empty);
if (trimWhiteSpace)
{
return(TrimWhiteSpace(HtmlEncoder.DecodeValue(str)));
}
return(HtmlEncoder.DecodeValue(str));
}
/// <summary>
/// This will parse a string containing HTML and will produce a domain tree.
/// </summary>
/// <param name="html">The HTML to be parsed</param>
/// <returns>A tree representing the elements</returns>
public HtmlNodeCollection Parse(string html)
{
HtmlNodeCollection nodes = new HtmlNodeCollection(null);
html = PreprocessScript(html, "script");
html = PreprocessScript(html, "style");
html = RemoveComments(html);
html = RemoveSGMLComments(html);
StringCollection tokens = GetTokens(html);
int index = 0;
HtmlElement element = null;
while (index < tokens.Count)
{
if ("<".Equals(tokens[index]))
{
// Read open tag
index++;
if (index >= tokens.Count)
{
break;
}
string tag_name = tokens[index];
index++;
element = new HtmlElement(tag_name);
// read the attributes and values
while (index < tokens.Count && !">".Equals(tokens[index]) && !"/>".Equals(tokens[index]))
{
string attribute_name = tokens[index];
index++;
if (index < tokens.Count && "=".Equals(tokens[index]))
{
index++;
string attribute_value;
if (index < tokens.Count)
{
attribute_value = tokens[index];
}
else
{
attribute_value = null;
}
index++;
HtmlAttribute attribute = new HtmlAttribute(attribute_name, HtmlEncoder.DecodeValue(attribute_value));
element.Attributes.Add(attribute);
}
else if (index < tokens.Count)
{
// Null-value attribute
HtmlAttribute attribute = new HtmlAttribute(attribute_name, null);
element.Attributes.Add(attribute);
}
}
nodes.Add(element);
if (index < tokens.Count && "/>".Equals(tokens[index]))
{
element.IsTerminated = true;
index++;
element = null;
}
else if (index < tokens.Count && ">".Equals(tokens[index]))
{
index++;
}
}
else if (">".Equals(tokens[index]))
{
index++;
}
else if ("</".Equals(tokens[index]))
{
// Read close tag
index++;
if (index >= tokens.Count)
{
break;
}
string tag_name = tokens[index];
index++;
int open_index = FindTagOpenNodeIndex(nodes, tag_name);
if (open_index != -1)
{
MoveNodesDown(ref nodes, open_index + 1, (HtmlElement)nodes[open_index]);
}
else
{
// Er, there is a close tag without an opening tag!!
}
// Skip to the end of this tag
while (index < tokens.Count && !">".Equals(tokens[index]))
{
index++;
}
if (index < tokens.Count && ">".Equals(tokens[index]))
{
index++;
}
element = null;
}
else
{
// Read text
string value = tokens[index];
if (mRemoveEmptyElementText)
{
value = RemoveWhitespace(value);
}
value = DecodeScript(value);
if (mRemoveEmptyElementText && value.Length == 0)
{
// We do nothing
}
else
{
if (!(element != null && element.NoEscaping))
{
value = HtmlEncoder.DecodeValue(value);
}
HtmlText node = new HtmlText(value);
nodes.Add(node);
}
index++;
}
}
return(nodes);
}
public HtmlNodeCollection Parse(string html)
{
HtmlNodeCollection nodes = new HtmlNodeCollection((HtmlElement)null);
html = this.PreprocessScript(html, "script");
html = this.PreprocessScript(html, "style");
html = this.RemoveComments(html);
html = this.RemoveSGMLComments(html);
StringCollection tokens = this.GetTokens(html);
int index1 = 0;
HtmlElement htmlElement = (HtmlElement)null;
while (index1 < tokens.Count)
{
if ("<".Equals(tokens[index1]))
{
int index2 = index1 + 1;
if (index2 >= tokens.Count)
{
return(nodes);
}
string name1 = tokens[index2];
index1 = index2 + 1;
htmlElement = new HtmlElement(name1);
while (index1 < tokens.Count && !">".Equals(tokens[index1]) && !"/>".Equals(tokens[index1]))
{
string name2 = tokens[index1];
++index1;
if (index1 < tokens.Count && "=".Equals(tokens[index1]))
{
int index3 = index1 + 1;
string str = index3 >= tokens.Count ? (string)null : tokens[index3];
index1 = index3 + 1;
HtmlAttribute attribute = new HtmlAttribute(name2, HtmlEncoder.DecodeValue(str));
htmlElement.Attributes.Add(attribute);
}
else if (index1 < tokens.Count)
{
HtmlAttribute attribute = new HtmlAttribute(name2, (string)null);
htmlElement.Attributes.Add(attribute);
}
}
nodes.Add((HtmlNode)htmlElement);
if (index1 < tokens.Count && "/>".Equals(tokens[index1]))
{
htmlElement.IsTerminated = true;
++index1;
htmlElement = (HtmlElement)null;
}
else if (index1 < tokens.Count && ">".Equals(tokens[index1]))
{
++index1;
}
}
else if (">".Equals(tokens[index1]))
{
++index1;
}
else if ("</".Equals(tokens[index1]))
{
int index2 = index1 + 1;
if (index2 >= tokens.Count)
{
return(nodes);
}
string name = tokens[index2];
index1 = index2 + 1;
int tagOpenNodeIndex = this.FindTagOpenNodeIndex(nodes, name);
if (tagOpenNodeIndex != -1)
{
this.MoveNodesDown(ref nodes, tagOpenNodeIndex + 1, (HtmlElement)nodes[tagOpenNodeIndex]);
}
while (index1 < tokens.Count && !">".Equals(tokens[index1]))
{
++index1;
}
if (index1 < tokens.Count && ">".Equals(tokens[index1]))
{
++index1;
}
htmlElement = (HtmlElement)null;
}
else
{
string str = tokens[index1];
if (this.mRemoveEmptyElementText)
{
str = this.RemoveWhitespace(str);
}
string text = HtmlParser.DecodeScript(str);
if (!this.mRemoveEmptyElementText || text.Length != 0)
{
if (htmlElement == null || !htmlElement.NoEscaping)
{
text = HtmlEncoder.DecodeValue(text);
}
HtmlText htmlText = new HtmlText(text);
nodes.Add((HtmlNode)htmlText);
}
++index1;
}
}
return(nodes);
}
public HtmlNodeCollection Parse(string html)
{
HtmlNodeCollection htmlNodeCollection = new HtmlNodeCollection(null);
html = this.PreprocessScript(html, "script");
html = this.PreprocessScript(html, "style");
html = this.RemoveComments(html);
html = this.RemoveSGMLComments(html);
StringCollection tokens = this.GetTokens(html);
int i = 0;
HtmlElement htmlElement = null;
HtmlNodeCollection result;
while (i < tokens.Count)
{
if ("<".Equals(tokens[i]))
{
i++;
if (i >= tokens.Count)
{
result = htmlNodeCollection;
return(result);
}
string name = tokens[i];
i++;
htmlElement = new HtmlElement(name);
while (i < tokens.Count && !">".Equals(tokens[i]) && !"/>".Equals(tokens[i]))
{
string name2 = tokens[i];
i++;
if (i < tokens.Count && "=".Equals(tokens[i]))
{
i++;
string value;
if (i < tokens.Count)
{
value = tokens[i];
}
else
{
value = null;
}
i++;
HtmlAttribute attribute = new HtmlAttribute(name2, HtmlEncoder.DecodeValue(value));
htmlElement.Attributes.Add(attribute);
}
else if (i < tokens.Count)
{
HtmlAttribute attribute = new HtmlAttribute(name2, null);
htmlElement.Attributes.Add(attribute);
}
}
htmlNodeCollection.Add(htmlElement);
if (i < tokens.Count && "/>".Equals(tokens[i]))
{
htmlElement.IsTerminated = true;
i++;
htmlElement = null;
}
else if (i < tokens.Count && ">".Equals(tokens[i]))
{
i++;
}
}
else if (">".Equals(tokens[i]))
{
i++;
}
else if ("</".Equals(tokens[i]))
{
i++;
if (i >= tokens.Count)
{
result = htmlNodeCollection;
return(result);
}
string name = tokens[i];
i++;
int num = this.FindTagOpenNodeIndex(htmlNodeCollection, name);
if (num != -1)
{
this.MoveNodesDown(ref htmlNodeCollection, num + 1, (HtmlElement)htmlNodeCollection[num]);
}
while (i < tokens.Count && !">".Equals(tokens[i]))
{
i++;
}
if (i < tokens.Count && ">".Equals(tokens[i]))
{
i++;
}
htmlElement = null;
}
else
{
string text = tokens[i];
if (this.mRemoveEmptyElementText)
{
text = this.RemoveWhitespace(text);
}
text = HtmlParser.DecodeScript(text);
if (!this.mRemoveEmptyElementText || text.Length != 0)
{
if (htmlElement == null || !htmlElement.NoEscaping)
{
text = HtmlEncoder.DecodeValue(text);
}
HtmlText node = new HtmlText(text);
htmlNodeCollection.Add(node);
}
i++;
}
}
result = htmlNodeCollection;
return(result);
}