Markup spec/BNF/Special block

Special blocks are things like itemised lists starting with *; they can only be specified at the start of a line and usually run until the end of the line (indicated by a line terminator sequence like CR LF or LF).

<special-block> ::=
	<horizontal-rule> |
	<heading> |
	<list-item> |
	<table> |
	<space-block> |
	...

The ellipsis needs to be filled in.

A horizontal rule is specified by 4 or more dashes. It is translated to an ‎<hr /> element.

<horizontal-rule> ::= "----" [<dashes>] [<inline-text>] <newline>
<dashes> ::= "-" [<dashes>]

If the <inline-text> is present, it is not wrapped in a ‎<p> element.

A level-N heading is translated to an ‎<hN> element.

<heading> ::=
	<level-6-heading> |
	<level-5-heading> |
	<level-4-heading> |
	<level-3-heading> |
	<level-2-heading> |
	<level-1-heading>
<level-6-heading> ::= "======" <inline-text> "======" <space-tabs> <newline>
<level-5-heading> ::= "=====" <inline-text> "=====" <space-tabs> <newline>
<level-4-heading> ::= "====" <inline-text> "====" <space-tabs> <newline>
<level-3-heading> ::= "===" <inline-text> "===" <space-tabs> <newline>
<level-2-heading> ::= "==" <inline-text> "==" <space-tabs> <newline>
<level-1-heading> ::= "=" <inline-text> "=" <space-tabs> <newline>

The alternatives in the <heading> rule need to be tried from top to bottom.

• An unterminated heading tag is treated as normal text.
• Unbalanced tags are treated as the shorter of the two tags (i.e. ==== heading == renders as the level-2 heading == heading ==)
• More than 6 = signs are treated as 6, with the extra symbols being included in the header.

<list-item> ::= <indent-item> | <enumerated-item> | <bullet-item> 
<indent-item> ::= ":" [(<list-item> | <item-body>)]
<enumerated-item> ::= "#" [(<list-item> | <item-body>)]
<bullet-item> ::= "*" [(<list-item> | <item-body>)]
<item-body> ::= <defined-term> | [<whitespace>] <inline-text>
<defined-term> ::= ";" <text> [(<definition>)]
<definition> ::= ":" <inline-text>

• <indent-item> and <definition> are translated to a ‎<dd> element, wrapped in a ‎<dl>
• A <bullet-item> is translated to a ‎<li> element wrapped in a ‎<ul>
• An <enumerated-item> is translated to a ‎<li> element wrapped in a ‎<ol>
• A <defined-term> is translated to a ‎<dt> element wrapped in a ‎<dl>

• The grouping of successive list items cannot be captured in EBNF. The simplest approach would appear to be a second pass whereby successive pairings of close/open list are eliminated. For example, <ol><li>Foo</li><s></ol><ol></s><li>Boo</li></ol> would be rewritten as <ol><li>Foo</li><li>Boo</li></ol>
• <list-item> and <defined-term> are obviously matched in preference to <inline-text>. The user has to insert whitespace in order to get inline-text starting with #, ;, * or :.
• The current parser accepts a wider range of syntax than the above, allowing other list items to appear after a definition list (;). This appears to be arbitrary, unpredictable and not particularly useful. See phab:T13894.

The following block is written in extended BNF (EBNF) syntax.

Table = "{|", [" ", TableParameters], NewLine, TableFirstRow, "|}";
TableFirstRow = TableColumnLine, NewLine | TableColumnMultiLine | TableRow;
TableRow = "|-", [CSS], NewLine, TableColumn, [TableRow];
TableColumn = TableColumnLine | TableColumnMultiLine;
TableColumnLine = "|", InlineText, ["|", TableColumnLine];
TableColumnMultiLine = "|", [TableCellParameters, "|"], AnyText, NewLine, [TableColumnMultiLine];
TableParameters = CSS | ? HTML table attributes ?;
TableCellParameter = CSS | ? HTML cell attributes ?;

Starting a line with a space creates a preformatted block of text similar to using ‎<pre>. The main difference is that the contained text is still parsed and rendered normally.

<space-block> ::= " " <inline-text> <newline> [{<space-block-2>}]
<space-block-2> ::= " " [<inline-text>] <newline>

• The block is surrounded with ‎<pre>. Whitespace and newlines are preserved literally.
• Note that the first line of a space block must have text in it. Subsequent lines can be composed of just spaces.

• Parser