if not modules then modules = { } end modules ['m-markdown'] = {
version = 1.002,
comment = "companion to m-markdown.mkiv",
author = "Hans Hagen, PRAGMA-ADE, Hasselt NL",
copyright = "see below",
license = "see context related readme files"
}
--[[
Copyright (C) 2009 John MacFarlane / Khaled Hosny / Hans Hagen
The main parser is derived from the lunamark parser written by John MacFarlane. You
can download lunamark from:
http://github.com/jgm/lunamark.git
Khaled Hosny provided the context writer for lunamark and that was used as starting
point for the mapping. The original code can be fetched from the above location.
While playing with the original code I got the feeling that lpeg could perform better.
The slowdown was due to the fact that the parser's lpeg was reconstructed each time a
nested parse was needed. After changing that code a bit I could bring down parsing of
some test code from 2 seconds to less than 0.1 second so I decided to stick to this
parser instead of writing my own. After all, the peg code looks pretty impressive and
visiting Johns pandoc pages is worth the effort:
http://johnmacfarlane.net/pandoc/
The code here is mostly meant for processing snippets embedded in a context
documents and is no replacement for pandoc at all. Therefore an alternative is to use
pandoc in combination with Aditya's filter module.
As I changed (and optimized) the original code, it will be clear that all errors
are mine. Eventually I might also adapt the parser code a bit more. When I ran into of
closure stack limitations I decided to flatten the code. The following implementation
seems to be a couple of hundred times faster than what I started with which is not that
bad.
This is a second rewrite. The mentioned speed gain largely depended on the kind of
content: blocks, references and items can be rather demanding. Also, There were
some limitations with respect to the captures. So, table storage has been removed in
favor of strings, and nesting has been simplified. The first example at the end of this
file now takes .33 seconds for 567KB code (resulting in over 1MB) so we're getting there.
There will be a third rewrite eventually.
]]--
-- todo: we have better quote and tag scanners in ctx
-- todo: provide an xhtml mapping
-- todo: add a couple of extensions
-- todo: check patches to the real peg
local type, next, tonumber = type, next, tonumber
local lower, upper, gsub, rep, gmatch, format, length = string.lower, string.upper, string.gsub, string.rep, string.gmatch, string.format, string.len
local concat = table.concat
local P, R, S, V, C, Ct, Cg, Cb, Cmt, Cc, Cf, Cs = lpeg.P, lpeg.R, lpeg.S, lpeg.V, lpeg.C, lpeg.Ct, lpeg.Cg, lpeg.Cb, lpeg.Cmt, lpeg.Cc, lpeg.Cf, lpeg.Cs
local lpegmatch = lpeg.match
local utfbyte, utfchar = utf.byte, utf.char
moduledata = moduledata or { }
moduledata.markdown = moduledata.markdown or { }
local markdown = moduledata.markdown
local nofruns, nofbytes, nofhtmlblobs = 0, 0, 0
---------------------------------------------------------------------------------------------
local nestedparser
local syntax
nestedparser = function(str) return lpegmatch(syntax,str) end
---------------------------------------------------------------------------------------------
local asterisk = P("*")
local dash = P("-")
local plus = P("+")
local underscore = P("_")
local period = P(".")
local hash = P("#")
local ampersand = P("&")
local backtick = P("`")
local less = P("<")
local more = P(">")
local space = P(" ")
local squote = P("'")
local dquote = P('"')
local lparent = P("(")
local rparent = P(")")
local lbracket = P("[")
local rbracket = P("]")
local slash = P("/")
local equal = P("=")
local colon = P(":")
local semicolon = P(";")
local exclamation = P("!")
local digit = R("09")
local hexdigit = R("09","af","AF")
local alphanumeric = R("AZ","az","09")
local doubleasterisks = P("**")
local doubleunderscores = P("__")
local fourspaces = P(" ")
local any = P(1)
local always = P("")
local tab = P("\t")
local spacechar = S("\t ")
local spacing = S(" \n\r\t")
local newline = P("\r")^-1 * P("\n")
local spaceornewline = spacechar + newline
local nonspacechar = any - spaceornewline
local optionalspace = spacechar^0
local spaces = spacechar^1
local eof = - any
local nonindentspace = space^-3
local blankline = optionalspace * C(newline)
local blanklines = blankline^0
local skipblanklines = (optionalspace * newline)^0
local linechar = P(1 - newline)
local indent = fourspaces + (nonindentspace * tab) / ""
local indentedline = indent /"" * C(linechar^1 * (newline + eof))
local optionallyindentedline = indent^-1 /"" * C(linechar^1 * (newline + eof))
local spnl = optionalspace * (newline * optionalspace)^-1
local specialchar = S("*_`*&[]
-- [3]:http://example.com/ (Optional Title Here)
-- [2]: http://example.com/ 'Optional Title Here'
-- [a]: http://example.com/ "Optional *oeps* Title Here"
-- ]]
--
-- local linktest = [[
-- [This link] (http://example.net/)
-- [an example] (http://example.com/ "Title")
-- [an example][1]
-- [an example] [2]
-- ]]
--
-- lpeg.match((define_reference_parser+1)^0,reftest)
--
-- inspect(references)
--
-- lpeg.match((direct_link_parser/print + indirect_link_parser/print + 1)^0,linktest)
---------------------------------------------------------------------------------------------
local blocktags = table.tohash {
"address", "blockquote" , "center", "dir", "div", "p", "pre",
"li", "ol", "ul", "dl", "dd",
"form", "fieldset", "isindex", "menu", "noframes", "frameset",
"h1", "h2", "h3", "h4", "h5", "h6",
"hr", "ht", "script", "noscript",
"table", "tbody", "tfoot", "thead", "th", "td", "tr",
}
----- htmlattributevalue = squote * C((any - (blankline + squote))^0) * squote
----- + dquote * C((any - (blankline + dquote))^0) * dquote
----- + (any - S("\t >"))^1 -- any - tab - space - more
----- htmlattribute = (alphanumeric + S("_-"))^1 * spnl * (equal * spnl * htmlattributevalue)^-1 * spnl
----- htmlcomment = P(""))^0 * P("-->")
----- htmltag = less * spnl * slash^-1 * alphanumeric^1 * spnl * htmlattribute^0 * slash^-1 * spnl * more
-----
----- blocktag = Cmt(C(alphanumeric^1), function(s,i,a) return blocktags[lower(a)] and i, a end)
-----
----- openblocktag = less * Cg(blocktag, "opentag") * spnl * htmlattribute^0 * more
----- closeblocktag = less * slash * Cmt(C(alphanumeric^1) * Cb("opentag"), function(s,i,a,b) return lower(a) == lower(b) and i end) * spnl * more
----- selfclosingblocktag = less * blocktag * spnl * htmlattribute^0 * slash * more
-----
----- displayhtml = Cs { "HtmlBlock",
----- InBlockTags = openblocktag * (V("HtmlBlock") + (any - closeblocktag))^0 * closeblocktag,
----- HtmlBlock = C(V("InBlockTags") + selfclosingblocktag + htmlcomment),
----- }
-----
----- inlinehtml = Cs(htmlcomment + htmltag)
-- There is no reason to support crappy html, so we expect proper attributes.
local htmlattributevalue = squote * C((any - (blankline + squote))^0) * squote
+ dquote * C((any - (blankline + dquote))^0) * dquote
local htmlattribute = (alphanumeric + S("_-"))^1 * spnl * equal * spnl * htmlattributevalue * spnl
local htmlcomment = P(""))^0 * P("-->")
local htmlinstruction = P("") * (any - P("?>" ))^0 * P("?>" )
-- We don't care too much about matching elements and there is no reason why display elements could not
-- have inline elements so the above should be patched then. Well, markdown mixed with html is not meant
-- for anything else than webpages anyway.
local blocktag = Cmt(C(alphanumeric^1), function(s,i,a) return blocktags[lower(a)] and i, a end)
local openelement = less * alphanumeric^1 * spnl * htmlattribute^0 * more
local closeelement = less * slash * alphanumeric^1 * spnl * more
local emptyelement = less * alphanumeric^1 * spnl * htmlattribute^0 * slash * more
local displaytext = (any - less)^1
local inlinetext = displaytext / nestedparser
local displayhtml = #(less * blocktag * spnl * htmlattribute^0 * more)
* Cs { "HtmlBlock",
InBlockTags = openelement * (V("HtmlBlock") + displaytext)^0 * closeelement,
HtmlBlock = (V("InBlockTags") + emptyelement + htmlcomment + htmlinstruction),
}
local inlinehtml = Cs { "HtmlBlock",
InBlockTags = openelement * (V("HtmlBlock") + inlinetext)^0 * closeelement,
HtmlBlock = (V("InBlockTags") + emptyelement + htmlcomment + htmlinstruction),
}
---------------------------------------------------------------------------------------------
local hexentity = ampersand * hash * S("Xx") * C(hexdigit ^1) * semicolon
local decentity = ampersand * hash * C(digit ^1) * semicolon
local tagentity = ampersand * C(alphanumeric^1) * semicolon
---------------------------------------------------------------------------------------------
-- --[[
local escaped = {
["{" ] = "",
["}" ] = "",
["$" ] = "",
["&" ] = "",
["#" ] = "",
["~" ] = "",
["|" ] = "",
["%%"] = "",
["\\"] = "",
}
for k, v in next, escaped do
escaped[k] = "\\char" .. utfbyte(k) .. "{}"
end
local function c_string(s) -- has to be done more often
return (gsub(s,".",escaped))
end
local c_linebreak = "\\crlf\n" -- is this ok?
local c_space = " "
local function c_paragraph(c)
return c .. "\n\n" -- { "\\startparagraph ", c, " \\stopparagraph\n" }
end
local function listitem(c)
return format("\n\\startitem\n%s\n\\stopitem\n",nestedparser(c))
end
local function c_tightbulletlist(c)
return format("\n\\startmarkdownitemize[packed]\n%s\\stopmarkdownitemize\n",c)
end
local function c_loosebulletlist(c)
return format("\n\\startmarkdownitemize\n\\stopmarkdownitemize\n",c)
end
local function c_tightorderedlist(c)
return format("\n\\startmarkdownitemize[n,packed]\n%s\\stopmarkdownitemize\n",c)
end
local function c_looseorderedlist(c)
return format("\n\\startmarkdownitemize[n]\n%s\\stopmarkdownitemize\n",c)
end
local function c_inline_html(content)
nofhtmlblobs = nofhtmlblobs + 1
return format("\\markdowninlinehtml{%s}",content)
end
local function c_display_html(content)
nofhtmlblobs = nofhtmlblobs + 1
return format("\\startmarkdowndisplayhtml\n%s\n\\stopmarkdowndisplayhtml",content)
end
local function c_emphasis(c)
return format("\\markdownemphasis{%s}",c)
end
local function c_strong(c)
return format("\\markdownstrong{%s}",c)
end
local function c_blockquote(c)
return format("\\startmarkdownblockquote\n%s\\stopmarkdownblockquote\n",nestedparser(c))
end
local function c_verbatim(c)
return format("\\startmarkdowntyping\n%s\\stopmarkdowntyping\n",c)
end
local function c_code(c)
return format("\\markdowntype{%s}",c)
end
local levels = { "", "", "", "", "", "" }
local function c_start_document()
levels = { "", "", "", "", "", "" }
return ""
end
local function c_stop_document()
return concat(levels,"\n") or ""
end
local function c_heading(level,c)
if level > #levels then
level = #levels
end
local finish = concat(levels,"\n",level) or ""
for i=level+1,#levels do
levels[i] = ""
end
levels[level] = "\\stopstructurelevel"
return format("%s\\startstructurelevel[markdown][title={%s}]\n",finish,c)
end
local function c_hrule()
return "\\markdownrule\n"
end
local function c_link(lab,src,tit)
return format("\\goto{%s}[url(%s)]",nestedparser(lab),src)
end
local function c_image(lab,src,tit)
return format("\\externalfigure[%s]",src)
end
local function c_email_link(address)
return format("\\goto{%s}[url(mailto:%s)]",c_string(address),address)
end
local function c_url_link(url)
return format("\\goto{%s}[url(%s)]",c_string(url),url)
end
local function f_heading(c,n)
return c_heading(n,c)
end
local function c_hex_entity(s)
return utfchar(tonumber(s,16))
end
local function c_dec_entity(s)
return utfchar(tonumber(s))
end
local function c_tag_entity(s)
return s -- we can use the default resolver
end
--]]
---------------------------------------------------------------------------------------------
--[[
local escaped = {
["<"] = "<",
[">"] = ">",
["&"] = "&",
['"'] = """,
}
local function c_string(s) -- has to be done more often
return (gsub(s,".",escaped))
end
local c_linebreak = "
"
local c_space = " "
local function c_paragraph(c)
return format("
%s
\n", c) end local function listitem(c) return format("\n%s\n",nestedparser(c)) end local function c_verbatim(c) return format("
%s
",c)
end
local function c_code(c)
return format("%s
",c)
end
local c_start_document = ""
local c_stop_document = ""
local function c_heading(level,c)
return format("PRE PRE PRE* Test ** Test * Test1 * Test2 * Test Test ==== > test > test **123** *123* > test `code` test Test ==== > test > test > test test oeps more code code oeps [an example][a] [an example] [2] [a]: http://example.com/ "Optional *oeps* Title Here" [2]: http://example.com/ 'Optional Title Here' [3]: http://example.com/ (Optional Title Here) [an example][a] [an example] [2] [an [tricky] example](http://example.com/ "Title") [This **xx** link](http://example.net/) ]] -- This snippet takes some 4 seconds in the original parser (the one that is -- a bit clearer from the perspective of grammars but somewhat messy with -- respect to the captures. In the above parser it takes .1 second. Also, -- in the later case only memory is the limit. local two = [[ Test ==== * Test ** Test * Test ** Test * Test Test ==== > test > test > test test Test ==== > test > test > test test ]] local function test(str) local n = 1 -- 000 local t = os.clock() local one = convert(str) -- print("runtime",1,#str,#one,os.clock()-t) str = string.rep(str,n) local t = os.clock() local two = convert(str) print(two) -- print("runtime",n,#str,#two,os.clock()-t) -- print(format("==============\n%s\n==============",one)) end -- test(one) -- test(two) -- test(io.read("*all")) end