private static IEnumerable <HTMLTag> ParseTags(string html, int htmlStart, int htmlEnd)
{
while (htmlStart < htmlEnd)
{
int pos = IndexOf(html, htmlStart, htmlEnd, '<');
if (pos == -1)
{
yield break;
}
HTMLTag tag = new HTMLTag();
tag.Offset = pos;
htmlStart = pos + 1;
tag.IsEnd = StartsWith(html, htmlStart, htmlEnd, '/');
if (StartsWithLetter(html, tag.IsEnd ? (htmlStart + 1) : htmlStart, htmlEnd))
{
// Parse tag name
if (tag.IsEnd)
{
htmlStart += 1;
}
pos = IndexOfAny(html, htmlStart, htmlEnd, true, '/', '>');
if (pos == -1)
{
yield break;
}
tag.Name = GetSectionLower(html, htmlStart, pos);
htmlStart = pos;
// Parse attributes
bool isTagComplete = false;
do
{
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd))
{
htmlStart++;
}
tag.IsSelfClosing = StartsWith(html, htmlStart, htmlEnd, '/');
if (tag.IsSelfClosing)
{
htmlStart += 1;
}
if (StartsWith(html, htmlStart, htmlEnd, '>'))
{
htmlStart += 1;
isTagComplete = true;
}
else if (tag.IsSelfClosing)
{
}
else
{
HTMLAttribute attribute = new HTMLAttribute();
attribute.Offset = htmlStart;
// Parse attribute name
pos = IndexOfAny(html, htmlStart + 1, htmlEnd, true, '=', '/', '>');
if (pos == -1)
{
yield break;
}
attribute.Name = GetSectionLower(html, htmlStart, pos);
htmlStart = pos;
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd))
{
htmlStart++;
}
if (StartsWith(html, htmlStart, htmlEnd, '='))
{
// Parse attribute value
htmlStart += 1;
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd))
{
htmlStart++;
}
if (StartsWithAny(html, htmlStart, htmlEnd, '"', '\''))
{
char quoteChar = html[htmlStart];
htmlStart += 1;
pos = IndexOf(html, htmlStart, htmlEnd, quoteChar);
if (pos == -1)
{
yield break;
}
attribute.Value = GetSection(html, htmlStart, pos);
htmlStart = pos + 1;
}
else
{
pos = IndexOfAny(html, htmlStart, htmlEnd, true, '>');
if (pos == -1)
{
yield break;
}
attribute.Value = GetSection(html, htmlStart, pos);
htmlStart = pos;
}
}
else
{
attribute.Value = String.Empty;
}
attribute.Length = htmlStart - attribute.Offset;
if (tag.GetAttribute(attribute.Name) == null)
{
tag.Attributes.Add(attribute);
}
}
} while (!isTagComplete);
tag.Length = htmlStart - tag.Offset;
// Yield result
yield return(tag);
// Skip contents of special tags whose contents are to be treated as raw text
if (!tag.IsEnd && !tag.IsSelfClosing && tag.NameEqualsAny("script", "style", "title", "textarea"))
{
bool foundEndTag = false;
do
{
pos = IndexOf(html, htmlStart, htmlEnd, '<');
if (pos == -1)
{
yield break;
}
htmlStart = pos + 1;
string endTagText = "/" + tag.Name;
if (StartsWith(html, htmlStart, htmlEnd, endTagText, true) &&
(StartsWithWhiteSpace(html, htmlStart + endTagText.Length, htmlEnd) ||
StartsWithAny(html, htmlStart + endTagText.Length, htmlEnd, '/', '>')))
{
htmlStart -= 1;
foundEndTag = true;
}
} while (!foundEndTag);
}
}
else if (StartsWith(html, htmlStart, htmlEnd, "!--", false) && !StartsWith(html, htmlStart + 3, htmlEnd, '>'))
{
// Skip comment
htmlStart += 3;
bool foundEnd = false;
do
{
pos = IndexOf(html, htmlStart, htmlEnd, '-');
if (pos == -1)
{
yield break;
}
htmlStart = pos + 1;
if (StartsWith(html, htmlStart, htmlEnd, "->", false))
{
htmlStart += 2;
foundEnd = true;
}
else if (StartsWith(html, htmlStart, htmlEnd, "-!>", false))
{
htmlStart += 3;
foundEnd = true;
}
} while (!foundEnd);
}
else if (StartsWithAny(html, htmlStart, htmlEnd, '?', '/', '!'))
{
// Skip bogus comment or DOCTYPE
htmlStart += 1;
pos = IndexOf(html, htmlStart, htmlEnd, '>');
if (pos == -1)
{
yield break;
}
htmlStart = pos + 1;
}
}
}
private static IEnumerable<HTMLTag> ParseTags(string html, int htmlStart, int htmlEnd)
{
while (htmlStart < htmlEnd) {
int pos = IndexOf(html, htmlStart, htmlEnd, '<');
if (pos == -1) yield break;
HTMLTag tag = new HTMLTag();
tag.Offset = pos;
htmlStart = pos + 1;
tag.IsEnd = StartsWith(html, htmlStart, htmlEnd, '/');
if (StartsWithLetter(html, tag.IsEnd ? (htmlStart + 1) : htmlStart, htmlEnd)) {
// Parse tag name
if (tag.IsEnd) htmlStart += 1;
pos = IndexOfAny(html, htmlStart, htmlEnd, true, '/', '>');
if (pos == -1) yield break;
tag.Name = GetSectionLower(html, htmlStart, pos);
htmlStart = pos;
// Parse attributes
bool isTagComplete = false;
do {
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd)) htmlStart++;
tag.IsSelfClosing = StartsWith(html, htmlStart, htmlEnd, '/');
if (tag.IsSelfClosing) htmlStart += 1;
if (StartsWith(html, htmlStart, htmlEnd, '>')) {
htmlStart += 1;
isTagComplete = true;
}
else if (tag.IsSelfClosing) { }
else {
HTMLAttribute attribute = new HTMLAttribute();
attribute.Offset = htmlStart;
// Parse attribute name
pos = IndexOfAny(html, htmlStart + 1, htmlEnd, true, '=', '/', '>');
if (pos == -1) yield break;
attribute.Name = GetSectionLower(html, htmlStart, pos);
htmlStart = pos;
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd)) htmlStart++;
if (StartsWith(html, htmlStart, htmlEnd, '=')) {
// Parse attribute value
htmlStart += 1;
while (StartsWithWhiteSpace(html, htmlStart, htmlEnd)) htmlStart++;
if (StartsWithAny(html, htmlStart, htmlEnd, '"', '\'')) {
char quoteChar = html[htmlStart];
htmlStart += 1;
pos = IndexOf(html, htmlStart, htmlEnd, quoteChar);
if (pos == -1) yield break;
attribute.Value = GetSection(html, htmlStart, pos);
htmlStart = pos + 1;
}
else {
pos = IndexOfAny(html, htmlStart, htmlEnd, true, '>');
if (pos == -1) yield break;
attribute.Value = GetSection(html, htmlStart, pos);
htmlStart = pos;
}
}
else {
attribute.Value = String.Empty;
}
attribute.Length = htmlStart - attribute.Offset;
if (tag.GetAttribute(attribute.Name) == null) {
tag.Attributes.Add(attribute);
}
}
} while (!isTagComplete);
tag.Length = htmlStart - tag.Offset;
// Yield result
yield return tag;
// Skip contents of special tags whose contents are to be treated as raw text
if (!tag.IsEnd && !tag.IsSelfClosing && tag.NameEqualsAny("script", "style", "title", "textarea")) {
bool foundEndTag = false;
do {
pos = IndexOf(html, htmlStart, htmlEnd, '<');
if (pos == -1) yield break;
htmlStart = pos + 1;
string endTagText = "/" + tag.Name;
if (StartsWith(html, htmlStart, htmlEnd, endTagText, true) &&
(StartsWithWhiteSpace(html, htmlStart + endTagText.Length, htmlEnd) ||
StartsWithAny(html, htmlStart + endTagText.Length, htmlEnd, '/', '>')))
{
htmlStart -= 1;
foundEndTag = true;
}
} while (!foundEndTag);
}
}
else if (StartsWith(html, htmlStart, htmlEnd, "!--", false) && !StartsWith(html, htmlStart + 3, htmlEnd, '>')) {
// Skip comment
htmlStart += 3;
bool foundEnd = false;
do {
pos = IndexOf(html, htmlStart, htmlEnd, '-');
if (pos == -1) yield break;
htmlStart = pos + 1;
if (StartsWith(html, htmlStart, htmlEnd, "->", false)) {
htmlStart += 2;
foundEnd = true;
}
else if (StartsWith(html, htmlStart, htmlEnd, "-!>", false)) {
htmlStart += 3;
foundEnd = true;
}
} while (!foundEnd);
}
else if (StartsWithAny(html, htmlStart, htmlEnd, '?', '/', '!')) {
// Skip bogus comment or DOCTYPE
htmlStart += 1;
pos = IndexOf(html, htmlStart, htmlEnd, '>');
if (pos == -1) yield break;
htmlStart = pos + 1;
}
}
}
