/// <summary>
/// Parses a PARAGRPAH_CLOSE .
/// </summary>
/// <param name="lastNode">The lastest parsed node.</param>
/// <returns>The extra paragraph, or <see cref="EMPTY_LINE_NODE"/>. If parsing attempt failed, <c>null</c>.</returns>
private LineNode ParseLineEnd(LineNode lastNode)
{
Debug.Assert(lastNode != null);
var unclosedParagraph = lastNode as Paragraph;
if (unclosedParagraph != null && !unclosedParagraph.Compact)
{
unclosedParagraph = null;
}
// 2 line breaks (\n\n) or \n Terminator closes the paragraph,
// so do a look-ahead here. Note that a \n will be consumed in ParseWikitext .
// Note that this look-ahead will also bypass the \n terminator defined in WIKITEXT
// For the last non-empty line
// TERM Terminators
// PC Compact/unclosed paragraph
// P Closed paragraph
// abc TERM PC[|abc|]
// abc\n TERM P[|abc|]
// abc\n\s*?\n TERM P[|abc|]PC[||]
// Note that MediaWiki editor will automatically trim the trailing whitespaces,
// leaving a \n after the content. This one \n will be removed when the page is transcluded.
var lastLinePosition = linePosition;
// Here we consume a \n without fallback.
if (ConsumeToken(@"\n") == null)
{
return(null);
}
ParseStart();
// Whitespaces between 2 \n, assuming there's a second \n or TERM after trailingWs
var trailingWs = ConsumeToken(@"[\f\r\t\v\x85\p{Z}]+");
if (unclosedParagraph != null)
{
// We're going to consume another \n or TERM to close the paragraph.
// Already consumed a \n, attempt to consume another \n
var trailingWsEndsAt = linePosition;
if (ConsumeToken(@"\n") != null)
{
// Close the last paragraph.
unclosedParagraph.AppendWithLineInfo("\n" + trailingWs,
// don't forget the position of leading '\n'
CurrentContext.StartingLineNumber - 1, lastLinePosition,
CurrentContext.StartingLineNumber, trailingWsEndsAt);
// 2 Line breaks received.
// Check for the special case. Note here TERM excludes \n
if (NeedsTerminate(Terminator.Get(@"\n")))
{
// This is a special case.
// abc \n trailingWs \n TERM --> P[|abc\ntrailingWs|]PC[||]
// ^ We are here.
// When the function returns, WIKITEXT parsing will stop
// because a TERM will be received.
// We need to correct this.
var anotherparagraph = new Paragraph();
anotherparagraph.SetLineInfo(lineNumber, linePosition, lineNumber, linePosition);
return(ParseSuccessful(anotherparagraph, false));
}
// The last paragraph will be closed now.
return(ParseSuccessful(EMPTY_LINE_NODE, false));
}
// The attempt to consume the 2nd \n failed.
if (NeedsTerminate())
{
// abc \n trailingWs TERM P[|abc|]
// ^ We are here.
// If we need to terminate, then close the last paragraph.
unclosedParagraph.AppendWithLineInfo("\n" + trailingWs,
// don't forget the position of leading '\n'
CurrentContext.StartingLineNumber - 1, lastLinePosition,
lineNumber, linePosition);
return(ParseSuccessful(EMPTY_LINE_NODE, false));
}
// The last paragraph is still not closed (i.e. compact paragraph).
// (A)
// Note here we have still consumed the first '\n', while the last paragraph has no trailing '\n'.
// For continued PlainText, we will add a '\n' in ParseCompactParagraph.
// Add an empty node so ParseCompactParagraph can add a '\n' with LineInfo.
unclosedParagraph.AppendWithLineInfo("", CurrentContext.StartingLineNumber - 1, lastLinePosition,
CurrentContext.StartingLineNumber - 1, lastLinePosition);
// Fallback so we can either continue parsing PlainText,
// or discover the next, for example, Heading, and leave the last paragraph compact.
Fallback();
return(EMPTY_LINE_NODE);
}
else
{
// Last node cannot be a closed paragraph.
// It can't because ParseLineEnd is invoked immediately after a last node is parsed,
// and only ParseLineEnd can close a paragraph.
Debug.Assert(!(lastNode is Paragraph), "Last node cannot be a closed paragraph.");
// Rather, last node is LINE node of other type (LIST_ITEM/HEADING).
// Remember we've already consumed a '\n' , and the spaces after it.
// The situation here is just like the "special case" mentioned above.
if (NeedsTerminate(Terminator.Get(@"\n")))
{
// abc \n WHITE_SPACE TERM --> [|abc|] PC[|WHITE_SPACE|]
// ^ CurCntxt ^ We are here now.
// Note here TERM excludes \n
var anotherparagraph = new Paragraph();
if (trailingWs != null)
{
var pt = new PlainText(trailingWs);
// Actually the same as what we do in ParseSuccessful for PlainText.
pt.SetLineInfo(CurrentContext.StartingLineNumber, CurrentContext.StartingLinePosition,
lineNumber, linePosition);
anotherparagraph.Inlines.Add(pt);
}
return(ParseSuccessful(anotherparagraph));
}
}
// abc \n def
// That's not the end of a prargraph. Fallback to before the 1st \n .
// Note here we have already consumed a \n .
Fallback();
return(EMPTY_LINE_NODE);
}
/// <summary>
/// HEADING
/// </summary>
private Heading ParseHeading()
{
// Look ahead to determine the maximum level, assuming the line is a valid heading.
var prefix = LookAheadToken("={1,6}");
if (prefix == null)
{
return(null);
}
// Note that here we require all the headings terminate with \n or EOF, so this won't be recognized
// {{{ARG|== Default Heading ==}}}\n
// But this will work
// {{{ARG|== Default Heading ==
// }}}
// Note that this should be recognized as heading
// == {{T|
// arg1}} ==
// Test different levels of heading
for (var level = prefix.Length; level > 0; level--)
{
var barExpr = "={" + level + "}";
// We use an early-stopping matching pattern
// E.g. for ==abc=={{def}}=={{ghi}}
// the first call to ParseRun will stop at =={{
// we need to continue parsing, resulting in a list of segments
// abc, {{def}}, {{ghi}}
var headingTerminator = barExpr + "(?!=)";
ParseStart(headingTerminator, false); // <-- A
var temp = ConsumeToken(barExpr);
Debug.Assert(temp != null);
var node = new Heading();
var parsedSegments = new List <IInlineContainer>();
while (true)
{
ParseStart(); // <-- B
var segment = new Run();
if (!ParseRun(RunParsingMode.Run, segment, true) &&
LookAheadToken(headingTerminator) == null)
{
// No more content to parse, and ParseRun stopped by
// a terminator that is not a heading terminator
// Stop and analyze
Fallback();
break;
}
if (ConsumeToken(barExpr) == null)
{
// The segment has been parsed, but it's terminated not by "==="
// We treat the last segment as suffix
// Stop and analyze
node.Suffix = segment;
Accept();
break;
}
// Put the run segment into the list.
parsedSegments.Add(segment);
}
if (node.Suffix != null &&
node.Suffix.Inlines.OfType <PlainText>().Any(pt => !string.IsNullOrWhiteSpace(pt.Content)))
{
// There shouldn't be non-whitespace plaintext after the heading
goto FAIL_CLEANUP;
}
node.Level = level;
// Concatenate segments, adding "===" where needed.
for (int i = 0; i < parsedSegments.Count; i++)
{
node.Inlines.AddFrom(parsedSegments[i].Inlines);
if (i < parsedSegments.Count - 1)
{
var si = (IWikitextLineInfo)parsedSegments[i + 1];
var bar = new PlainText(new string('=', level));
bar.SetLineInfo(si.StartLineNumber, si.StartLinePosition - level,
si.StartLineNumber, si.StartLinePosition);
node.Inlines.Add(bar);
}
}
if (node.Inlines.Count == 0)
{
// There should be something as heading content
goto FAIL_CLEANUP;
}
// Move forward
// -- B
for (int i = 0; i < parsedSegments.Count; i++)
{
Accept();
}
return(ParseSuccessful(node)); // <-- A
FAIL_CLEANUP:
// -- B
for (int i = 0; i < parsedSegments.Count; i++)
{
Fallback();
}
Fallback(); // <-- A
}
// Failed (E.g. ^=== Title )
return(null);
}