/// <summary>
/// Parses a markdown document.
/// </summary>
/// <param name="markdown"> The markdown text. </param>
/// <param name="start"> The position to start parsing. </param>
/// <param name="end"> The position to stop parsing. </param>
/// <param name="quoteDepth"> The current nesting level for block quoting. </param>
/// <param name="actualEnd"> Set to the position at which parsing ended. This can be
/// different from <paramref name="end"/> when the parser is being called recursively.
/// </param>
/// <returns> A list of parsed blocks. </returns>
internal static List <MarkdownBlock> Parse(string markdown, int start, int end, int quoteDepth, out int actualEnd)
{
// We need to parse out the list of blocks.
// Some blocks need to start on a new paragraph (code, lists and tables) while other
// blocks can start on any line (headers, horizontal rules and quotes).
// Text that is outside of any other block becomes a paragraph.
var blocks = new List <MarkdownBlock>();
int startOfLine = start;
bool lineStartsNewParagraph = true;
var paragraphText = new StringBuilder();
// These are needed to parse underline-style header blocks.
int previousStartOfLine = start;
int previousEndOfLine = start;
// Go line by line.
while (startOfLine < end)
{
// Find the first non-whitespace character.
int nonSpacePos = startOfLine;
char nonSpaceChar = '\0';
int realStartOfLine = startOfLine; // i.e. including quotes.
int expectedQuotesRemaining = quoteDepth;
while (true)
{
while (nonSpacePos < end)
{
char c = markdown[nonSpacePos];
if (c == '\r' || c == '\n')
{
// The line is either entirely whitespace, or is empty.
break;
}
if (c != ' ' && c != '\t')
{
// The line has content.
nonSpaceChar = c;
break;
}
nonSpacePos++;
}
// When parsing blocks in a blockquote context, we need to count the number of
// quote characters ('>'). If there are less than expected AND this is the
// start of a new paragraph, then stop parsing.
if (expectedQuotesRemaining == 0)
{
break;
}
if (nonSpaceChar == '>')
{
// Expected block quote characters should be ignored.
expectedQuotesRemaining--;
nonSpacePos++;
nonSpaceChar = '\0';
startOfLine = nonSpacePos;
// Ignore the first space after the quote character, if there is one.
if (startOfLine < end && markdown[startOfLine] == ' ')
{
startOfLine++;
nonSpacePos++;
}
}
else
{
// There were less block quote characters than expected.
// But it doesn't matter if this is not the start of a new paragraph.
if (!lineStartsNewParagraph || nonSpaceChar == '\0')
{
break;
}
// This must be the end of the blockquote. End the current paragraph, if any.
actualEnd = previousEndOfLine;
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
}
return(blocks);
}
}
// Find the end of the current line.
int startOfNextLine;
int endOfLine = Helpers.Common.FindNextSingleNewLine(markdown, nonSpacePos, end, out startOfNextLine);
if (nonSpaceChar == '\0')
{
// The line is empty or nothing but whitespace.
lineStartsNewParagraph = true;
// End the current paragraph.
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
paragraphText.Clear();
}
}
else
{
// This is a header if the line starts with a hash character,
// or if the line starts with '-' or a '=' character and has no other characters.
// Or a quote if the line starts with a greater than character (optionally preceded by whitespace).
// Or a horizontal rule if the line contains nothing but 3 '*', '-' or '_' characters (with optional whitespace).
MarkdownBlock newBlockElement = null;
if (lineStartsNewParagraph)
{
// Some block elements must start on a new paragraph (tables, lists and code).
int endOfBlock = startOfNextLine;
if (nonSpacePos > startOfLine || nonSpaceChar == '`')
{
newBlockElement = CodeBlock.Parse(markdown, realStartOfLine, end, quoteDepth, out endOfBlock);
}
if (newBlockElement != null)
{
startOfNextLine = endOfBlock;
}
}
// This check needs to go after the code block check.
if (newBlockElement == null && nonSpaceChar == '[')
{
newBlockElement = LinkReferenceBlock.Parse(markdown, startOfLine, endOfLine);
}
// Block elements start new paragraphs.
lineStartsNewParagraph = newBlockElement != null;
if (newBlockElement == null)
{
// The line contains paragraph text.
if (paragraphText.Length > 0)
{
paragraphText.Append("\n");
}
// Add the last paragraph if we are at the end of the input text.
if (startOfNextLine >= end)
{
if (paragraphText.Length == 0)
{
// Optimize for single line paragraphs.
blocks.Add(ParagraphBlock.Parse(markdown.Substring(startOfLine, endOfLine - startOfLine)));
}
else
{
// Slow path.
paragraphText.Append(markdown.Substring(startOfLine, endOfLine - startOfLine));
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
}
}
else
{
paragraphText.Append(markdown.Substring(startOfLine, endOfLine - startOfLine));
}
}
else
{
// The line contained a block. End the current paragraph, if any.
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
paragraphText.Clear();
}
blocks.Add(newBlockElement);
}
}
// Repeat.
previousStartOfLine = startOfLine;
previousEndOfLine = endOfLine;
startOfLine = startOfNextLine;
}
actualEnd = startOfLine;
return(blocks);
}
/// <summary>
/// Parses a markdown document.
/// </summary>
/// <param name="markdown"> The markdown text. </param>
/// <param name="start"> The position to start parsing. </param>
/// <param name="end"> The position to stop parsing. </param>
/// <param name="quoteDepth"> The current nesting level for block quoting. </param>
/// <param name="actualEnd"> Set to the position at which parsing ended. This can be
/// different from <paramref name="end"/> when the parser is being called recursively.
/// </param>
/// <returns> A list of parsed blocks. </returns>
internal static List <MarkdownBlock> Parse(string markdown, int start, int end, int quoteDepth, out int actualEnd)
{
// We need to parse out the list of blocks.
// Some blocks need to start on a new paragraph (code, lists and tables) while other
// blocks can start on any line (headers, horizontal rules and quotes).
// Text that is outside of any other block becomes a paragraph.
var blocks = new List <MarkdownBlock>();
int startOfLine = start;
bool lineStartsNewParagraph = true;
var paragraphText = new StringBuilder();
// These are needed to parse underline-style header blocks.
int previousRealtStartOfLine = start;
int previousStartOfLine = start;
int previousEndOfLine = start;
// Go line by line.
while (startOfLine < end)
{
// Find the first non-whitespace character.
int nonSpacePos = startOfLine;
char nonSpaceChar = '\0';
int realStartOfLine = startOfLine; // i.e. including quotes.
int expectedQuotesRemaining = quoteDepth;
while (true)
{
while (nonSpacePos < end)
{
char c = markdown[nonSpacePos];
if (c == '\r' || c == '\n')
{
// The line is either entirely whitespace, or is empty.
break;
}
if (c != ' ' && c != '\t')
{
// The line has content.
nonSpaceChar = c;
break;
}
nonSpacePos++;
}
// When parsing blocks in a blockquote context, we need to count the number of
// quote characters ('>'). If there are less than expected AND this is the
// start of a new paragraph, then stop parsing.
if (expectedQuotesRemaining == 0)
{
break;
}
if (nonSpaceChar == '>')
{
// Expected block quote characters should be ignored.
expectedQuotesRemaining--;
nonSpacePos++;
nonSpaceChar = '\0';
startOfLine = nonSpacePos;
// Ignore the first space after the quote character, if there is one.
if (startOfLine < end && markdown[startOfLine] == ' ')
{
startOfLine++;
nonSpacePos++;
}
}
else
{
int lastIndentation = 0;
string lastline = null;
// Determines how many Quote levels were in the last line.
if (realStartOfLine > 0)
{
lastline = markdown.Substring(previousRealtStartOfLine, previousEndOfLine - previousRealtStartOfLine);
lastIndentation = lastline.Count(c => c == '>');
}
var currentEndOfLine = Common.FindNextSingleNewLine(markdown, nonSpacePos, end, out _);
var currentline = markdown.Substring(realStartOfLine, currentEndOfLine - realStartOfLine);
var currentIndentation = currentline.Count(c => c == '>');
var firstChar = markdown[realStartOfLine];
// This is a quote that doesn't start with a Quote marker, but carries on from the last line.
if (lastIndentation == 1)
{
if (nonSpaceChar != '\0' && firstChar != '>')
{
break;
}
}
// Collapse down a level of quotes if the current indentation is greater than the last indentation.
// Only if the last indentation is greater than 1, and the current indentation is greater than 0
if (lastIndentation > 1 && currentIndentation > 0 && currentIndentation < lastIndentation)
{
break;
}
// This must be the end of the blockquote. End the current paragraph, if any.
actualEnd = realStartOfLine;
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
}
return(blocks);
}
}
// Find the end of the current line.
int startOfNextLine;
int endOfLine = Common.FindNextSingleNewLine(markdown, nonSpacePos, end, out startOfNextLine);
if (nonSpaceChar == '\0')
{
// The line is empty or nothing but whitespace.
lineStartsNewParagraph = true;
// End the current paragraph.
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
paragraphText.Clear();
}
}
else
{
// This is a header if the line starts with a hash character,
// or if the line starts with '-' or a '=' character and has no other characters.
// Or a quote if the line starts with a greater than character (optionally preceded by whitespace).
// Or a horizontal rule if the line contains nothing but 3 '*', '-' or '_' characters (with optional whitespace).
MarkdownBlock newBlockElement = null;
if (nonSpaceChar == '#' && nonSpacePos == startOfLine)
{
// Hash-prefixed header.
newBlockElement = HeaderBlock.ParseHashPrefixedHeader(markdown, startOfLine, endOfLine);
}
else if ((nonSpaceChar == '-' || nonSpaceChar == '=') && nonSpacePos == startOfLine && paragraphText.Length > 0)
{
// Underline style header. These are weird because you don't know you've
// got one until you've gone past it.
// Note: we intentionally deviate from reddit here in that we only
// recognize this type of header if the previous line is part of a
// paragraph. For example if you have this, the header at the bottom is
// ignored:
// a|b
// -|-
// 1|2
// ===
newBlockElement = HeaderBlock.ParseUnderlineStyleHeader(markdown, previousStartOfLine, previousEndOfLine, startOfLine, endOfLine);
if (newBlockElement != null)
{
// We're going to have to remove the header text from the pending
// paragraph by prematurely ending the current paragraph.
// We already made sure that there is a paragraph in progress.
paragraphText.Length = paragraphText.Length - (previousEndOfLine - previousStartOfLine);
}
}
// These characters overlap with the underline-style header - this check should go after that one.
if (newBlockElement == null && (nonSpaceChar == '*' || nonSpaceChar == '-' || nonSpaceChar == '_'))
{
newBlockElement = HorizontalRuleBlock.Parse(markdown, startOfLine, endOfLine);
}
if (newBlockElement == null && lineStartsNewParagraph)
{
// Some block elements must start on a new paragraph (tables, lists and code).
int endOfBlock = startOfNextLine;
if (nonSpaceChar == '*' || nonSpaceChar == '+' || nonSpaceChar == '-' || (nonSpaceChar >= '0' && nonSpaceChar <= '9'))
{
newBlockElement = ListBlock.Parse(markdown, realStartOfLine, end, quoteDepth, out endOfBlock);
}
if (newBlockElement == null && (nonSpacePos > startOfLine || nonSpaceChar == '`'))
{
newBlockElement = CodeBlock.Parse(markdown, realStartOfLine, end, quoteDepth, out endOfBlock);
}
if (newBlockElement == null)
{
newBlockElement = TableBlock.Parse(markdown, realStartOfLine, endOfLine, end, quoteDepth, out endOfBlock);
}
if (newBlockElement != null)
{
startOfNextLine = endOfBlock;
}
}
// This check needs to go after the code block check.
if (newBlockElement == null && nonSpaceChar == '>')
{
newBlockElement = QuoteBlock.Parse(markdown, realStartOfLine, end, quoteDepth, out startOfNextLine);
}
// This check needs to go after the code block check.
if (newBlockElement == null && nonSpaceChar == '[')
{
newBlockElement = LinkReferenceBlock.Parse(markdown, startOfLine, endOfLine);
}
// Block elements start new paragraphs.
lineStartsNewParagraph = newBlockElement != null;
if (newBlockElement == null)
{
// The line contains paragraph text.
if (paragraphText.Length > 0)
{
// If the previous two characters were both spaces, then append a line break.
if (paragraphText.Length > 2 && paragraphText[paragraphText.Length - 1] == ' ' && paragraphText[paragraphText.Length - 2] == ' ')
{
// Replace the two spaces with a line break.
paragraphText[paragraphText.Length - 2] = '\r';
paragraphText[paragraphText.Length - 1] = '\n';
}
else
{
paragraphText.Append(" ");
}
}
// Add the last paragraph if we are at the end of the input text.
if (startOfNextLine >= end)
{
if (paragraphText.Length == 0)
{
// Optimize for single line paragraphs.
blocks.Add(ParagraphBlock.Parse(markdown.Substring(startOfLine, endOfLine - startOfLine)));
}
else
{
// Slow path.
paragraphText.Append(markdown.Substring(startOfLine, endOfLine - startOfLine));
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
}
}
else
{
paragraphText.Append(markdown.Substring(startOfLine, endOfLine - startOfLine));
}
}
else
{
// The line contained a block. End the current paragraph, if any.
if (paragraphText.Length > 0)
{
blocks.Add(ParagraphBlock.Parse(paragraphText.ToString()));
paragraphText.Clear();
}
blocks.Add(newBlockElement);
}
}
// Repeat.
previousRealtStartOfLine = realStartOfLine;
previousStartOfLine = startOfLine;
previousEndOfLine = endOfLine;
startOfLine = startOfNextLine;
}
actualEnd = startOfLine;
return(blocks);
}
/// <summary>
/// Attempts to parse a reference e.g. "[example]: http://www.reddit.com 'title'".
/// </summary>
/// <param name="markdown"> The markdown text. </param>
/// <param name="start"> The location to start parsing. </param>
/// <param name="end"> The location to stop parsing. </param>
/// <returns> A parsed markdown link, or <c>null</c> if this is not a markdown link. </returns>
internal static LinkReferenceBlock Parse(string markdown, int start, int end)
{
// Expect a '[' character.
if (start >= end || markdown[start] != '[')
{
return(null);
}
// Find the ']' character
int pos = start + 1;
while (pos < end)
{
if (markdown[pos] == ']')
{
break;
}
pos++;
}
if (pos == end)
{
return(null);
}
// Extract the ID.
string id = markdown.Substring(start + 1, pos - (start + 1));
// Expect the ':' character.
pos++;
if (pos == end || markdown[pos] != ':')
{
return(null);
}
// Skip whitespace
pos++;
while (pos < end && ParseHelpers.IsMarkdownWhiteSpace(markdown[pos]))
{
pos++;
}
if (pos == end)
{
return(null);
}
// Extract the URL.
int urlStart = pos;
while (pos < end && !ParseHelpers.IsMarkdownWhiteSpace(markdown[pos]))
{
pos++;
}
string url = TextRunInline.ResolveEscapeSequences(markdown, urlStart, pos);
// Ignore leading '<' and trailing '>'.
url = url.TrimStart('<').TrimEnd('>');
// Skip whitespace.
pos++;
while (pos < end && ParseHelpers.IsMarkdownWhiteSpace(markdown[pos]))
{
pos++;
}
string tooltip = null;
if (pos < end)
{
// Extract the tooltip.
char tooltipEndChar;
switch (markdown[pos])
{
case '(':
tooltipEndChar = ')';
break;
case '"':
case '\'':
tooltipEndChar = markdown[pos];
break;
default:
return(null);
}
pos++;
int tooltipStart = pos;
while (pos < end && markdown[pos] != tooltipEndChar)
{
pos++;
}
if (pos == end)
{
return(null); // No end character.
}
tooltip = markdown.Substring(tooltipStart, pos - tooltipStart);
// Check there isn't any trailing text.
pos++;
while (pos < end && ParseHelpers.IsMarkdownWhiteSpace(markdown[pos]))
{
pos++;
}
if (pos < end)
{
return(null);
}
}
// We found something!
var result = new LinkReferenceBlock {
Id = id, Url = url, Tooltip = tooltip
};
return(result);
}
