/// <summary>
/// Parses an opening tag like '<div>'. Starts with the input stream
/// pointing to the opening < character.
/// </summary>
private Element ProcessTagOpenState(Element currentElement)
{
if (!_characterStream.MoveNext())
{
_document.ConformanceLevel *= 0.8f;
return(null);
}
switch (_characterStream.Current)
{
case '!':
_characterStream.State = HtmlStates.MarkupDeclarationOpen;
return(currentElement);
case '/':
_characterStream.State = HtmlStates.EndTagOpen;
return(currentElement);
}
var startPosition = _characterStream.CurrentPosition;
string tagName = null;
var selfClosing = false;
using (var nameBuffer = _stringBuilderFactory.Create(_maximumNameLength))
{
char?terminator;
_characterStream.State = HtmlStates.TagName;
if (_stringParser.TakeUntil(nameBuffer, _maximumNameLength, c => char.IsWhiteSpace(c) || c == '>' || c == '/', out terminator))
{
tagName = nameBuffer.ToString().ToLower();
if (!terminator.HasValue || terminator == '/')
{
_characterStream.State = HtmlStates.SelfClosingStartTag;
if (_stringParser.Peek() == '>')
{
_stringParser.TakeOne();
}
else
{
_document.ConformanceLevel *= 0.9f;
}
}
else if (terminator == '>')
{
_characterStream.State = _voidElements.Contains(tagName)
? HtmlStates.SelfClosingStartTag
: HtmlStates.Data;
}
else
{
_characterStream.State = HtmlStates.AttributeName;
selfClosing = _voidElements.Contains(tagName);
}
}
else
{
_document.ConformanceLevel *= 0.9f;
nameBuffer.Clear();
_stringParser.Take(nameBuffer, _maximumNameLength);
var buffer = nameBuffer.ToString().ToLower();
_characterStream.State = HtmlStates.Data;
foreach (var name in _allElements)
{
if (buffer.StartsWith(name))
{
nameBuffer.Clear();
_characterStream.Reset(startPosition);
_stringParser.Take(nameBuffer, name.Length);
tagName = nameBuffer.ToString();
_characterStream.State = HtmlStates.AttributeName;
selfClosing = _voidElements.Contains(tagName);
break;
}
}
if (tagName == null)
{
_document.ConformanceLevel *= 0.6f;
}
}
}
var attributes = _characterStream.State == HtmlStates.AttributeName ? ParseAttributes(selfClosing) : null;
if (currentElement != null)
{
var parentElement = currentElement;
switch (tagName)
{
case "html":
case "body":
case "form":
case "header":
case "footer":
// These elements are not parsed and contain no details. They are included in the output
// only as containers for their children
currentElement = new UnsupportedElement {
Attributes = attributes, SuppressOutput = false
};
break;
case "p":
case "li":
// These elements are treated as paragraphs in other markup formats, for
// example in markdown they will have a blank line above them to create a
// paragraph break.
currentElement = new ParagraphElement {
Attributes = attributes
};
break;
case "blockquote":
currentElement = new ParagraphElement
{
Attributes = attributes,
Styles = new Dictionary <string, string>
{
{ "margin-top", "10px" },
{ "margin-bottom", "10px" },
{ "margin-left", "50px" },
{ "padding-left", "15px" },
{ "border-left", "3px solid #ccc" }
}
};
break;
case "div":
// Divs are tricky because some pwople use them to group elements with similar
// style and other people used them instead of paragraphs. Since divs are by
// default block elements it makes more sense in most cases to treat them link
// paragraphs unless they have paraphraphs or other divs within them.
currentElement = new ContainerElement {
ContainerType = ContainerTypes.Division, Attributes = attributes
};
break;
case "span":
// These elements are treated as inline text. For example in markdown
// these are rendered without an extra blank line and are therefore rendered
// as part of the prior paragraph
currentElement = new SpanElement {
Attributes = attributes
};
break;
case "a":
// Anchor tags are a special case
if (attributes != null && attributes.ContainsKey("href"))
{
currentElement = new AnchorElement {
LinkAddress = attributes["href"]
}
}
;
else
{
currentElement = new UnsupportedElement {
Attributes = attributes
}
};
break;
case "iframe":
case "img":
// Image tags are a special case
if (attributes != null && attributes.ContainsKey("src"))
{
var alt = attributes.ContainsKey("alt") ? attributes["alt"] : null;
currentElement = new ImageElement {
LinkAddress = attributes["src"], AltText = alt
};
}
else
{
currentElement = new UnsupportedElement {
Attributes = attributes
};
}
break;
case "h1":
currentElement = new HeadingElement {
Level = 1
};
break;
case "h2":
currentElement = new HeadingElement {
Level = 2
};
break;
case "h3":
currentElement = new HeadingElement {
Level = 3
};
break;
case "h4":
currentElement = new HeadingElement {
Level = 4
};
break;
case "h5":
currentElement = new HeadingElement {
Level = 5
};
break;
case "h6":
currentElement = new HeadingElement {
Level = 6
};
break;
case "strong":
case "b":
// Bold is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "font-weight", "bold" }
}
};
break;
case "cite":
case "q":
case "i":
case "em":
// Italic is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "font-style", "italic" }
}
};
break;
case "u":
// Underline is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "text-decoration", "underline" }
}
};
break;
case "small":
// Small is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "font-size", "smaller" }
}
};
break;
case "sup":
// Superscript is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "vertical-align", "super" },
{ "font-size", "smaller" }
}
};
break;
case "sub":
// Subscript is represented as an inline style
currentElement = new FormattedElement
{
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string>
{
{ "vertical-align", "sub" },
{ "font-size", "smaller" }
}
};
break;
case "br":
currentElement = new BreakElement {
BreakType = BreakTypes.LineBreak
};
break;
case "hr":
currentElement = new BreakElement {
BreakType = BreakTypes.HorizontalRule
};
break;
case "ul":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.BulletList, Attributes = attributes
};
break;
case "ol":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.NumberedList, Attributes = attributes
};
break;
case "table":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.Table, Attributes = attributes
};
break;
case "tr":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.TableDataRow, Attributes = attributes
};
break;
case "th":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.TableHeaderRow, Attributes = attributes
};
break;
case "td":
currentElement = new ContainerElement {
ContainerType = ContainerTypes.TableDataCell, Attributes = attributes
};
break;
default:
// All other elements will be excluded from the output document, but will
// be parsed just so that we know where they and and the next valid element
// begins.
currentElement = new UnsupportedElement {
Attributes = attributes
};
break;
}
var styleElement = currentElement as IStyleElement;
if (styleElement != null && attributes != null)
{
if (attributes.ContainsKey("class"))
{
styleElement.ClassNames = attributes["class"];
attributes.Remove("class");
}
if (attributes.ContainsKey("style"))
{
if (styleElement.Styles == null)
{
styleElement.Styles = new Dictionary <string, string>();
}
var styles = attributes["style"].Split(';').Select(s => s.Trim()).Where(s => s.Length > 0);
foreach (var style in styles)
{
var colonPos = style.IndexOf(':');
if (colonPos > 0 && colonPos < style.Length - 1)
{
var name = style.Substring(0, colonPos).Trim().ToLower();
var value = style.Substring(colonPos + 1).Trim().ToLower();
if (!styleElement.Styles.ContainsKey(name))
{
styleElement.Styles[name] = value;
}
}
else
{
_document.ConformanceLevel *= 0.9f;
}
}
attributes.Remove("style");
}
}
currentElement.Name = tagName;
currentElement.Parent = parentElement;
if (parentElement.SuppressOutput)
{
currentElement.SuppressOutput = true;
}
if (parentElement.Children == null)
{
parentElement.Children = new List <IDocumentElement>();
}
parentElement.Children.Add(currentElement);
if (!Begin(currentElement))
{
return(null);
}
}
return(currentElement);
}
private void ParseText(string text)
{
var isBold = false;
var isItalic = false;
var isCode = false;
using (var buffer = _stringBuilderFactory.Create())
{
Action <IStringBuilder> flush = b =>
{
if (b.Length > 0)
{
PushElement(new RawTextElement {
Text = b.ToString()
});
PopElement();
b.Clear();
}
};
for (var i = 0; i < text.Length; i++)
{
var prior2 = i > 1 ? text[i - 2] : default(char);
var prior1 = i > 0 ? text[i - 1] : default(char);
var current = text[i];
var next = i < text.Length - 2 ? text[i + 1] : default(char);
if ((prior1 == '*' && current == '*') || (prior1 == '_' && current == '_'))
{
// Double asterix or double underline turns bold on/off
flush(buffer);
if (isBold)
{
PopElement();
isBold = false;
}
else
{
PushElement(new FormattedElement
{
Name = new String(current, 2),
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string> {
{ "font-weight", "bold" }
}
});
isBold = true;
}
}
else if ((prior1 == '*' || prior1 == '_') && (prior1 != prior2))
{
// Single asterix or single underline turns italic on/off
flush(buffer);
if (isItalic)
{
PopElement();
isItalic = false;
}
else
{
PushElement(new FormattedElement
{
Name = new String(prior1, 1),
ElementType = ElementTypes.InlineText,
Styles = new Dictionary <string, string> {
{ "font-style", "italic" }
}
});
isItalic = true;
}
buffer.Append(current);
}
else if (current == '*' || current == '_')
{
// When we see the first asterix or underscore, we don't know yet if this
// is going to be bold or italic unless this is the last character in the string
}
else if (current == '`')
{
// Backticks turn on/off code formatting
flush(buffer);
if (isCode)
{
PopElement();
isCode = false;
}
else
{
PushElement(new FormattedElement
{
Name = new String(current, 2),
ElementType = ElementTypes.InlineText,
ClassNames = "code"
});
isCode = true;
}
}
else if (current == '!')
{
// Can be an image link if it is followed by []
if (next != '[')
{
buffer.Append(current);
}
}
else if (current == '[')
{
// Open square bracket is the start of a hyperlink
if (text.Length < i + 3)
{
buffer.Append(current);
continue;
}
var firstCloseIndex = text.IndexOf(']', i + 1);
if (firstCloseIndex < 0)
{
buffer.Append(current);
continue;
}
AnchorElement anchor = null;
var title = text.Substring(i + 1, firstCloseIndex - i - 1);
var nextChar = text.Length > firstCloseIndex + 3 ? text[firstCloseIndex + 1] : default(char);
var isImageLink = prior1 == '!';
if (nextChar == '(')
{
// Url is inline with the link
var secondCloseIndex = text.IndexOf(')', firstCloseIndex + 2);
if (secondCloseIndex == -1)
{
secondCloseIndex = text.Length;
}
anchor = new AnchorElement
{
Name = "()",
LinkAddress = text.Substring(firstCloseIndex + 2, secondCloseIndex - firstCloseIndex - 2)
};
i = secondCloseIndex;
}
else if (nextChar == '[')
{
// Url is provided elsewhere in the document as a reference
var secondCloseIndex = text.IndexOf(']', firstCloseIndex + 2);
if (secondCloseIndex == -1)
{
secondCloseIndex = text.Length;
}
anchor = new AnchorElement
{
Name = "[]",
LinkAddress = text.Substring(firstCloseIndex + 2, secondCloseIndex - firstCloseIndex - 2)
};
_anchorsToFixup.Add(anchor);
i = secondCloseIndex;
}
else if (nextChar == ':')
{
var urlStartIndex = firstCloseIndex + 2;
while (char.IsWhiteSpace(text[urlStartIndex]))
{
urlStartIndex++;
}
var urlEndIndex = urlStartIndex + 1;
while (urlEndIndex < text.Length && !char.IsWhiteSpace(text[urlEndIndex]))
{
urlEndIndex++;
}
var url = text.Substring(urlStartIndex, urlEndIndex - urlStartIndex);
_references[title] = url;
i = urlEndIndex - 1;
}
else
{
// Self-referencing
anchor = new AnchorElement
{
Name = "[]",
LinkAddress = title
};
_anchorsToFixup.Add(anchor);
i = firstCloseIndex;
}
if (anchor != null)
{
flush(buffer);
if (isImageLink)
{
anchor.LinkType = LinkTypes.Image;
anchor.AltText = title;
PushElement(anchor);
PopElement(); // Pop the anchor to add it to its parent
}
else
{
PushElement(anchor);
PushElement(new RawTextElement {
Text = title
});
PopElement(); // Pop the raw text
PopElement(); // Pop the anchor
}
}
}
else if (current == '<')
{
// Angle brackets are the start of a hyperlink with no title
var closeIndex = text.IndexOf('>', i + 1);
if (closeIndex < 0)
{
buffer.Append(current);
continue;
}
var url = text.Substring(i + 1, closeIndex - i - 1);
flush(buffer);
PushElement(new AnchorElement {
Name = "<>", LinkAddress = url
});
PushElement(new RawTextElement {
Text = url
});
PopElement(); // Pop the raw text
PopElement(); // Pop the anchor
i = closeIndex;
}
else
{
buffer.Append(current);
}
}
flush(buffer);
if (isItalic) // If italic was opened but not closed then close it here
{
PopElement();
}
if (isBold) // If bold was opened but not closed then close it here
{
PopElement();
}
}
}
