public static void Tokenize(string htmlText, IWriter w)
{
var r = new CharReader(htmlText);
htmlText = null;
var sb = new StringBuilder();
var lastTagIx = -1;
string tagName = null;
string attrName = null;
var attrs = new List <Attr>();
char attrValQuotes;
r.Read();
textCur :;
if (!r.IsEof && (r.Chr != '<' || !r.Peek().IsHtmlTagBeginning()))
{
sb.Append(r.Chr);
r.Read();
goto textCur;
}
if (sb.Length != 0)
{
var text = sb.ToString();
sb.Clear();
w.Text(text);
}
if (r.IsEof)
{
goto eofCur;
}
lastTagIx = r.Ix;
r.Read();
tagNameCur :;
if (!r.IsEof && !r.Chr.IsHtmlWhiteSpace() && r.Chr != '>' && (r.Chr != '/' || r.Peek() != '>'))
{
sb.Append(r.Chr);
r.Read();
goto tagNameCur;
}
tagName = sb.ToString();
sb.Clear();
if (tagName.StartsWith("!--", StringComparison.Ordinal))
{
tagName = null;
goto commentsCur;
}
if (tagName == "!DOCTYPE")
{
tagName = null;
goto doctypeCur;
}
tagBodyCur :;
if (r.Chr.IsHtmlWhiteSpace())
{
r.Read();
goto tagBodyCur;
}
if (attrName != null && (r.Chr != '=' || attrName == "/"))
{
attrs.Add(new Attr(attrName, null));
attrName = null;
}
if (r.IsEof || r.Chr == '>')
{
if (r.Chr == '>')
{
r.Read();
}
goto tagDoneCur;
}
if (attrName == null)
{
goto attrNameCur;
}
r.Read();
goto attrPreValCur;
attrNameCur :;
if (!r.IsEof && !r.Chr.IsHtmlWhiteSpace() && r.Chr != '>' && (r.Chr != '/' || sb.Length == 0) && r.Chr != '=')
{
sb.Append(r.Chr);
r.Read();
goto attrNameCur;
}
attrName = sb.ToString();
sb.Clear();
goto tagBodyCur;
attrPreValCur :;
if (r.Chr.IsHtmlWhiteSpace())
{
r.Read();
goto attrPreValCur;
}
if (r.Chr == '"' || r.Chr == '\'')
{
attrValQuotes = r.Chr;
r.Read();
}
else
{
attrValQuotes = '\0';
}
attrValCur :;
if (!r.IsEof && (attrValQuotes != '\0' ? r.Chr != attrValQuotes : !r.Chr.IsHtmlWhiteSpace() && r.Chr != '>'))
{
sb.Append(r.Chr);
r.Read();
goto attrValCur;
}
if (!r.IsEof && attrValQuotes != '\0')
{
r.Read();
}
var attrVal = sb.ToString();
sb.Clear();
attrs.Add(new Attr(attrName, attrVal));
attrName = null;
goto tagBodyCur;
tagDoneCur :;
w.Tag(tagName, attrs.AsReadOnly());
attrs.Clear();
var tagNameSaved = tagName;
tagName = null;
if (r.IsEof || (true &&
!tagNameSaved.EqualsIgnoreCase("script") && !tagNameSaved.EqualsIgnoreCase("style") &&
!tagNameSaved.EqualsIgnoreCase("iframe") && !tagNameSaved.EqualsIgnoreCase("textarea") && !tagNameSaved.EqualsIgnoreCase("title")
))
{
goto textCur;
}
var fragmBeginIx = r.Ix;
var m = new Regex(string.Format(@"</{0}\b", tagNameSaved), RegexOptions.CultureInvariant | RegexOptions.IgnoreCase | RegexOptions.Singleline)
.Match(r.Text, fragmBeginIx);
var fragmEndIx = m.Success ? m.Index : r.Text.Length;
if (fragmEndIx == fragmBeginIx)
{
goto textCur;
}
r.SkipTo(fragmEndIx);
if (tagNameSaved.EqualsIgnoreCase("script") || tagNameSaved.EqualsIgnoreCase("style"))
{
w.Inlay(r.Text.Substring(fragmBeginIx, fragmEndIx - fragmBeginIx), tagNameSaved.EqualsIgnoreCase("script") ? HtmlInlayType.Script : HtmlInlayType.Style);
}
else
{
w.Text(r.Text.Substring(fragmBeginIx, fragmEndIx - fragmBeginIx));
}
goto textCur;
commentsCur :;
var commentsBeginIx = lastTagIx;
var commentsEndIx = r.Text.IndexOf("-->", commentsBeginIx + 4, StringComparison.Ordinal);
commentsEndIx = commentsEndIx >= 0 ? commentsEndIx + 3 : r.Text.Length;
if (!r.IsEof)
{
r.SkipTo(commentsEndIx);
}
w.Inlay(r.Text.Substring(commentsBeginIx, commentsEndIx - commentsBeginIx), HtmlInlayType.Comments);
goto textCur;
doctypeCur :;
var doctypeBeginIx = lastTagIx;
m = new Regex(@"<!DOCTYPE\s*(([^\s""'>]+|""[^""]*(""|$)|'[^']*('|$))\s*)*(>|$)", RegexOptions.CultureInvariant | RegexOptions.IgnoreCase | RegexOptions.Singleline)
.Match(r.Text, doctypeBeginIx);
Trace.Assert(m.Success);
var doctypeEndIx = m.Index + m.Length;
if (!r.IsEof)
{
r.SkipTo(doctypeEndIx);
}
w.Inlay(r.Text.Substring(doctypeBeginIx, doctypeEndIx - doctypeBeginIx), HtmlInlayType.Doctype);
goto textCur;
eofCur :;
r = null;
w.Eof();
}
public static string Decode(string textHtml)
{
var r = new CharReader(textHtml);
textHtml = null;
var sb = new StringBuilder();
var escIx = -1;
var escCode = -1;
r.Read();
textCur :;
if (!r.IsEof && r.Chr != '&')
{
sb.Append(r.Chr);
r.Read();
goto textCur;
}
if (r.IsEof)
{
goto eofCur;
}
escIx = r.Ix;
r.Read();
if (r.Chr.IsHtmlEscLit())
{
goto escLitCur;
}
if (r.Chr != '#')
{
goto escDoneCur;
}
r.Read();
if (r.Chr.IsHtmlEscNum())
{
goto escNumCur;
}
if (r.Chr != 'X' && r.Chr != 'x')
{
goto escDoneCur;
}
r.Read();
if (r.Chr.IsHtmlEscHex())
{
goto escHexCur;
}
goto escDoneCur;
escLitCur :;
if (r.Chr.IsHtmlEscLit())
{
r.Read();
goto escLitCur;
}
var escLit = r.Text.Substring(escIx + 1, r.Ix - (escIx + 1));
if (!HtmlCharCodes.CharNameToCode.TryGetValue(escLit, out escCode))
{
escCode = -1;
}
goto escDoneCur;
escNumCur :;
if (r.Chr.IsHtmlEscNum())
{
r.Read();
goto escNumCur;
}
var escNum = r.Text.Substring(escIx + 2, r.Ix - (escIx + 2));
if (!int.TryParse(escNum, NumberStyles.Integer, CultureInfo.InvariantCulture, out escCode) || escCode > 0xFFFD)
{
escCode = 0xFFFD;
}
goto escDoneCur;
escHexCur :;
if (r.Chr.IsHtmlEscHex())
{
r.Read();
goto escHexCur;
}
var escHex = r.Text.Substring(escIx + 3, r.Ix - (escIx + 3));
if (!int.TryParse(escHex, NumberStyles.HexNumber, CultureInfo.InvariantCulture, out escCode) || escCode > 0xFFFD)
{
escCode = 0xFFFD;
}
escDoneCur :;
if (escCode >= 0)
{
sb.Append((char)escCode);
escCode = -1;
if (r.Chr == ';')
{
r.Read();
}
escIx = -1;
}
if (escIx >= 0)
{
sb.Append(r.Text, escIx, r.Ix - escIx);
escIx = -1;
}
goto textCur;
eofCur :;
return(sb.ToString());
}
