#!/usr/bin/env lua

--[[
# markdown.lua -- version 0.13

<http://www.frykholm.se/files/markdown.lua>

**Author:** Niklas Frykholm, <niklas@frykholm.se>  
**Date:** 15 May 2006

This is an implementation of the popular text markup language Markdown in pure Lua.
Markdown can convert documents written in a simple and easy to read text format
to well-formatted HTML. For a more thourough description of Markdown and the Markdown
syntax, see <http://daringfireball.net/projects/markdown>.

The original Markdown source is written in Perl and makes heavy use of advanced
regular expression techniques (such as negative look-ahead, etc) which are not available
in Lua's simple regex engine. Therefore this Lua port has been rewritten from the ground
up. It is probably not completely bug free. If you notice any bugs, please report them to
me. A unit test that exposes the error is helpful.

## Usage

    require "markdown"
    markdown(source)

``markdown.lua`` exposes a single global function named ``markdown(s)`` which applies the
Markdown transformation to the specified string.

``markdown.lua`` can also be used directly from the command line:

	lua markdown.lua test.md

Creates a file ``test.html`` with the converted content of ``test.md``. Run:

    lua markdown.lua -h

For a description of the command-line options.

``markdown.lua`` uses the same license as Lua, the MIT license.

## License

Copyright &copy; 2006 Niklas Frykholm.

Permission is hereby granted, free of charge, to any person obtaining a copy of this
software and associated documentation files (the "Software"), to deal in the Software
without restriction, including without limitation the rights to use, copy, modify, merge,
publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons
to whom the Software is furnished to do so, subject to the following conditions: 

The above copyright notice and this permission notice shall be included in all copies
or substantial portions of the Software. 

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.

## Version history

-   **0.13** -- 12 Aug 2006
	-	Added -l option for including stylesheet inline in document.
	-	Fixed bug in -s flag.
	-	Fixed emphasis bug.
-	**0.12** -- 15 May 2006
	-	Fixed several bugs to comply with MarkdownTest 1.0 <http://six.pairlist.net/pipermail/markdown-discuss/2004-December/000909.html>
-	**0.11** -- 12 May 2006
	-	Fixed bug for escaping `*` and `_` inside code spans.
	-	Added license terms.
	-	Changed join() to table.concat().
-	**0.10** -- 3 May 2006
	-	Initial public release.

// Niklas
]]


-- Set up a table for holding local functions to avoid polluting the global namespace
local M = {}
local MT = {__index = _G}
setmetatable(M, MT)
setfenv(1, M)

----------------------------------------------------------------------
-- Utility functions
----------------------------------------------------------------------

-- Locks table t from changes, writes an error if someone attempts to change the table.
-- This is useful for detecting variables that have "accidently" been made global. Something
-- I tend to do all too much.
function lock(t)
	function lock_new_index(t, k, v)
		error("module has been locked -- " .. k .. " must be declared local", 2)
	end

	local mt = {__newindex = lock_new_index}
	if getmetatable(t) then mt.__index = getmetatable(t).__index end
	setmetatable(t, mt)
end

-- Returns the result of mapping the values in table t through the function f
function map(t, f)
	local out = {}
	for k,v in pairs(t) do out[k] = f(v,k) end
	return out
end

-- The identity function, useful as a placeholder.
function identity(text) return text end

-- Functional style if statement. (NOTE: no short circuit evaluation)
function iff(t, a, b) if t then return a else return b end end

-- Splits the text into an array of separate lines.
function split(text, sep)
	sep = sep or "\n"
	local lines = {}
	local pos = 1
	while true do
		local b,e = text:find(sep, pos)
		if not b then table.insert(lines, text:sub(pos)) break end
		table.insert(lines, text:sub(pos, b-1))
		pos = e + 1
	end
	return lines
end

-- Converts tabs to spaces
function detab(text)
	local tab_width = 4
	local function rep(match)
		local spaces = -match:len()
		while spaces<1 do spaces = spaces + tab_width end
		return match .. string.rep(" ", spaces)
	end
	text = text:gsub("([^\n]-)\t", rep)
	return text
end

-- Applies string.find for every pattern in the list and returns the first match
function find_first(s, patterns, index)
	local res = {}
	for _,p in ipairs(patterns) do
		local match = {s:find(p, index)}
		if #match>0 and (#res==0 or match[1] < res[1]) then res = match end
	end
	return unpack(res)
end

-- If a replacement array is specified, the range [start, stop] in the array is replaced
-- with the replacement array and the resulting array is returned. Without a replacement
-- array the section of the array between start and stop is returned.
function splice(array, start, stop, replacement)
	if replacement then
		local n = stop - start + 1
		while n > 0 do
			table.remove(array, start)
			n = n - 1
		end
		for i,v in ipairs(replacement) do
			table.insert(array, start, v)
		end
		return array
	else
		local res = {}
		for i = start,stop do
			table.insert(res, array[i])
		end
		return res
	end
end

-- Outdents the text one step.
function outdent(text)
	text = "\n" .. text
	text = text:gsub("\n  ? ? ?", "\n")
	text = text:sub(2)
	return text
end

-- Indents the text one step.
function indent(text)
	text = text:gsub("\n", "\n    ")
	return text
end

-- Does a simple tokenization of html data. Returns the data as a list of tokens. 
-- Each token is a table with a type field (which is either "tag" or "text") and
-- a text field (which contains the original token data).
function tokenize_html(html)
	local tokens = {}
	local pos = 1
	while true do
		local start = find_first(html, {"<!%-%-", "<[a-z/!$]", "<%?"}, pos)
		if not start then
			table.insert(tokens, {type="text", text=html:sub(pos)})
			break
		end
		if start ~= pos then table.insert(tokens, {type="text", text = html:sub(pos, start-1)}) end
		
		local _, stop
		if html:match("^<!%-%-", start) then
			_,stop = html:find("%-%->", start)
		elseif html:match("^<%?", start) then
			_,stop = html:find("?>", start)
		else
			_,stop = html:find("%b<>", start)
		end
		if not stop then error("Could not match html tags") end
		table.insert(tokens, {type="tag", text=html:sub(start, stop)})
		pos = stop + 1
	end
	return tokens
end

----------------------------------------------------------------------
-- Hash
----------------------------------------------------------------------

-- This is used to "hash" data into alphanumeric strings that are unique
-- in the document. (Note that this is not cryptographic hash, the hash
-- function is not one-way.) The hash procedure is used to protect parts
-- of the document from further processing.

local HASH = {
	-- Has the hash been inited.
	inited = false,
	
	-- The unique string prepended to all hash values. This is to ensure
	-- that hash values do not accidently coincide with an actual existing
	-- string in the document.
	identifier = "",
	
	-- Counter that counts up for each new hash instance.
	counter = 0,
	
	-- Hash table.
	table = {}
}

-- Inits hashing. Creates a hash_identifier that doesn't occur anywhere
-- in the text.
function init_hash(text)
	HASH.inited = true
	HASH.identifier = ""
	HASH.counter = 0
	HASH.table = {}
	
	local s = "HASH"
	local counter = 0
	local id
	while true do
		id  = s .. counter
		if not text:find(id, 1, true) then break end
		counter = counter + 1
	end
	HASH.identifier = id
end

-- Returns the hashed value for s.
function hash(s)
	assert(HASH.inited)
	if not HASH.table[s] then
		HASH.counter = HASH.counter + 1
		local id = HASH.identifier .. HASH.counter .. "X"
		HASH.table[s] = id
	end
	return HASH.table[s]
end

----------------------------------------------------------------------
-- Protection
----------------------------------------------------------------------

-- The protection module is used to "protect" parts of a document
-- so that they are not modified by subsequent processing steps. 
-- Protected parts are saved in a table for later unprotection

-- Protection data
local PD = {
	-- Saved blocks that have been converted
	blocks = {},

	-- Block level tags that will be protected
	tags = {"p", "div", "h1", "h2", "h3", "h4", "h5", "h6", "blockquote",
	"pre", "table", "dl", "ol", "ul", "script", "noscript", "form", "fieldset",
	"iframe", "math", "ins", "del"}
}

-- Pattern for matching a block tag that begins and ends in the leftmost
-- column and may contain indented subtags, i.e.
-- <div>
--    A nested block.
--    <div>
--        Nested data.
--     </div>
-- </div>
function block_pattern(tag)
	return "\n<" .. tag .. ".-\n</" .. tag .. ">[ \t]*\n"
end

-- Pattern for matching a block tag that begins and ends with a newline
function line_pattern(tag)
	return "\n<" .. tag .. ".-</" .. tag .. ">[ \t]*\n"
end

-- Protects the range of characters from start to stop in the text and
-- returns the protected string.
function protect_range(text, start, stop)
	local s = text:sub(start, stop)
	local h = hash(s)
	PD.blocks[h] = s
	text = text:sub(1,start) .. h .. text:sub(stop)
	return text
end

-- Protect every part of the text that matches any of the patterns. The first
-- matching pattern is protected first, etc.
function protect_matches(text, patterns)
	while true do
		local start, stop = find_first(text, patterns)
		if not start then break end
		text = protect_range(text, start, stop)
	end
	return text
end

-- Protects blocklevel tags in the specified text
function protect(text)
	-- First protect potentially nested block tags
	text = protect_matches(text, map(PD.tags, block_pattern))
	-- Then protect block tags at the line level.
	text = protect_matches(text, map(PD.tags, line_pattern))
	-- Protect <hr> and comment tags
	text = protect_matches(text, {"\n<hr[^>]->[ \t]*\n"})
	text = protect_matches(text, {"\n<!%-%-.-%-%->[ \t]*\n"})
	return text
end

-- Returns true if the string s is a hash resulting from protection
function is_protected(s)
	return PD.blocks[s]
end

-- Unprotects the specified text by expanding all the nonces
function unprotect(text)
	for k,v in pairs(PD.blocks) do
		text = text:gsub(k, v)
	end
	return text
end


----------------------------------------------------------------------
-- Block transform
----------------------------------------------------------------------

-- The block transform functions transform the text on the block level.
-- They work with the text as an array of lines rather than as individual
-- characters.

-- Returns true if the line is a ruler of (char) characters.
-- The line must contain at least three char characters and contain only spaces and
-- char characters.
function is_ruler_of(line, char)
	if not line:match("^[ %" .. char .. "]*$") then return false end
	if not line:match("%" .. char .. ".*%" .. char .. ".*%" .. char) then return false end
	return true
end

-- Identifies the block level formatting present in the line
function classify(line)
	local info = {line = line, text = line}
	
	if line:match("^    ") then
		info.type = "indented"
		info.outdented = line:sub(5)
		return info
	end
	
	for _,c in ipairs({'*', '-', '_', '='}) do
		if is_ruler_of(line, c) then
			info.type = "ruler"
			info.ruler_char = c
			return info
		end
	end
	
	if line == "" then
		info.type = "blank"
		return info
	end
	
	if line:match("^(#+)[ \t]*(.-)[ \t]*#*[ \t]*$") then
		local m1, m2 = line:match("^(#+)[ \t]*(.-)[ \t]*#*[ \t]*$")
		info.type = "header"
		info.level = m1:len()
		info.text = m2
		return info
	end
	
	if line:match("^ ? ? ?(%d+)%.[ \t]+(.+)") then
		local number, text = line:match("^ ? ? ?(%d+)%.[ \t]+(.+)")
		info.type = "list_item"
		info.list_type = "numeric"
		info.number = 0 + number
		info.text = text
		return info
	end
	
	if line:match("^ ? ? ?([%*%+%-])[ \t]+(.+)") then
		local bullet, text = line:match("^ ? ? ?([%*%+%-])[ \t]+(.+)")
		info.type = "list_item"
		info.list_type = "bullet"
		info.bullet = bullet
		info.text= text
		return info
	end
	
	if line:match("^>[ \t]?(.*)") then
		info.type = "blockquote"
		info.text = line:match("^>[ \t]?(.*)")
		return info
	end
	
	if is_protected(line) then
		info.type = "raw"
		info.html = unprotect(line)
		return info
	end
	
	info.type = "normal"
	return info
end

-- Find headers constisting of a normal line followed by a ruler and converts them to
-- header entries.
function headers(array)
	local i = 1
	while i <= #array - 1 do
		if array[i].type  == "normal" and array[i+1].type == "ruler" and 
			(array[i+1].ruler_char == "-" or array[i+1].ruler_char == "=") then
			local info = {line = array[i].line}
			info.text = info.line
			info.type = "header"
			info.level = iff(array[i+1].ruler_char == "=", 1, 2)
			table.remove(array, i+1)
			array[i] = info
		end
		i = i + 1
	end
	return array
end

-- Find list blocks and convert them to protected data blocks
function lists(array, sublist)
	local function process_list(arr)
		local function any_blanks(arr)
			for i = 1, #arr do
				if arr[i].type == "blank" then return true end
			end
			return false
		end
		
		local function split_list_items(arr)
			local acc = {arr[1]}
			local res = {}
			for i=2,#arr do
				if arr[i].type == "list_item" then
					table.insert(res, acc)
					acc = {arr[i]}
				else
					table.insert(acc, arr[i])
				end
			end
			table.insert(res, acc)
			return res
		end
		
		local function process_list_item(lines, block)
			while lines[#lines].type == "blank" do
				table.remove(lines)
			end
			
			local itemtext = lines[1].text
			for i=2,#lines do
				itemtext = itemtext .. "\n" .. outdent(lines[i].line)
			end
			if block then
				itemtext = block_transform(itemtext, true)
				return "    <li>" .. indent(itemtext) .. "</li>"
			else
				local lines = split(itemtext)
				lines = map(lines, classify)
				lines = lists(lines, true)
				lines = blocks_to_html(lines, true)
				itemtext = table.concat(lines, "\n")
				return "    <li>" .. indent(itemtext) .. "</li>"
			end
		end
		
		local block_list = any_blanks(arr)
		local items = split_list_items(arr)
		local out = ""
		for _, item in ipairs(items) do
			out = out .. process_list_item(item, block_list) .. "\n"
		end
		if arr[1].list_type == "numeric" then
			return "<ol>\n" .. out .. "</ol>"
		else
			return "<ul>\n" .. out .. "</ul>"
		end
	end
	
	-- Finds the range of lines composing the first list in the array. A list
	-- starts with (^ list_item) or (blank list_item) and ends with
	-- (blank* $) or (blank normal).
	--
	-- A sublist can start with just (list_item) does not need a blank...
	local function find_list(array, sublist)
		local function find_list_start(array, sublist)
			if array[1].type == "list_item" then return 1 end
			if sublist then
				for i = 1,#array do
					if array[i].type == "list_item" then return i end
				end
			else
				for i = 1, #array-1 do
					if array[i].type == "blank" and array[i+1].type == "list_item" then
						return i+1
					end
				end
			end
			return nil
		end
		local function find_list_end(array, start)
			local pos = #array
			for i = start, #array-1 do
				if array[i].type == "blank" and array[i+1].type ~= "list_item" 
					and array[i+1].type ~= "indented" and array[i+1].type ~= "blank" then
					pos = i-1
					break
				end
			end
			while pos > start and array[pos].type == "blank" do
				pos = pos - 1
			end
			return pos
		end
		
		local start = find_list_start(array, sublist)
		if not start then return nil end
		return start, find_list_end(array, start)
	end
	
	while true do
		local start, stop = find_list(array, sublist)
		if not start then break end
		local text = process_list(splice(array, start, stop))
		local info = {
			line = text,
			type = "raw",
			html = text
		}
		array = splice(array, start, stop, {info})
	end
	
	-- Convert any remaining list items to normal
	for _,line in ipairs(array) do
		if line.type == "list_item" then line.type = "normal" end
	end
	
	return array
end

-- Find and convert blockquote markers.
function blockquotes(lines)
	local function find_blockquote(lines)
		local start
		for i,line in ipairs(lines) do
			if line.type == "blockquote" then
				start = i
				break
			end
		end
		if not start then return nil end
		
		local stop = #lines
		for i = start+1, #lines do
			if lines[i].type == "blank" or lines[i].type == "blockquote" then
			elseif lines[i].type == "normal" then
				if lines[i-1].type == "blank" then stop = i-1 break end
			else
				stop = i-1 break
			end
		end
		while lines[stop].type == "blank" do stop = stop - 1 end
		return start, stop
	end
	
	local function process_blockquote(lines)
		local raw = lines[1].text
		for i = 2,#lines do
			raw = raw .. "\n" .. lines[i].text
		end
		local bt = block_transform(raw)
		if not bt:find("<pre>") then bt = indent(bt) end
		return "<blockquote>\n    " .. bt ..
			"\n</blockquote>"
	end
	
	while true do
		local start, stop = find_blockquote(lines)
		if not start then break end
		local text = process_blockquote(splice(lines, start, stop))
		local info = {
			line = text,
			type = "raw",
			html = text
		}
		lines = splice(lines, start, stop, {info})
	end
	return lines
end

-- Find and convert codeblocks.
function codeblocks(lines)
	local function find_codeblock(lines)
		local start
		for i,line in ipairs(lines) do
			if line.type == "indented" then start = i break end
		end
		if not start then return nil end
		
		local stop = #lines
		for i = start+1, #lines do
			if lines[i].type ~= "indented" and lines[i].type ~= "blank" then
				stop = i-1
				break
			end
		end
		while lines[stop].type == "blank" do stop = stop - 1 end
		return start, stop
	end
	
	local function process_codeblock(lines)
		local raw = detab(encode_code(outdent(lines[1].line)))
		for i = 2,#lines do
			raw = raw .. "\n" .. detab(encode_code(outdent(lines[i].line)))
		end
		return "<pre><code>" .. raw .. "\n</code></pre>"
	end
	
	while true do
		local start, stop = find_codeblock(lines)
		if not start then break end
		local text = process_codeblock(splice(lines, start, stop))
		local info = {
			line = text,
			type = "raw",
			html = text
		}
		lines = splice(lines, start, stop, {info})
	end
	return lines
end

-- Convert lines to html code
function blocks_to_html(lines, no_paragraphs)
	local out = {}
	local i = 1
	while i <= #lines do
		local line = lines[i]
		if line.type == "ruler" then
			table.insert(out, "<hr/>")
		elseif line.type == "raw" then
			table.insert(out, line.html)
		elseif line.type == "normal" then
			local s = line.line
			
			while i+1 <= #lines and lines[i+1].type == "normal" do
				i = i + 1
				s = s .. "\n" .. lines[i].line
			end
			
			if no_paragraphs then
				table.insert(out, span_transform(s))
			else
				table.insert(out, "<p>" .. span_transform(s) .. "</p>")
			end
		elseif line.type == "header" then
			local s = "<h" .. line.level .. ">" .. span_transform(line.text) .. "</h" .. line.level .. ">"
			table.insert(out, s)
		else
			table.insert(out, line.line)
		end
		i = i + 1
	end
	return out
end

-- Perform all the block level transforms
function block_transform(text, sublist)
	local lines = split(text)
	lines = map(lines, classify)
	lines = headers(lines)
	lines = lists(lines, sublist)
	lines = codeblocks(lines)
	lines = blockquotes(lines)
	lines = blocks_to_html(lines)
	local text = table.concat(lines, "\n")
	return text
end

-- Debug function for printing a line array to see the result
-- of partial transforms.
function print_lines(lines)
	for i, line in ipairs(lines) do
		print(i, line.type, line.text or line.line)
	end
end

----------------------------------------------------------------------
-- Span transform
----------------------------------------------------------------------

-- Functions for transforming the text at the span level.

-- These characters may need to be escaped because they have a special
-- meaning in markdown.
escape_chars = "'\\`*_{}[]()>#+-.!'"
escape_table = {}

function init_escape_table()
	for i = 1,#escape_chars do
		local c = escape_chars:sub(i,i)
		escape_table[c] = hash(c)
	end
end

-- Escape characters that should not be disturbed by markdown.
function escape_special_chars(text)
	local tokens = tokenize_html(text)
	
	local out = ""
	for _, token in ipairs(tokens) do
		local t = token.text
		if token.type == "tag" then
			-- In tags, encode * and _ so they don't conflict with their use in markdown.
			t = t:gsub("%*", escape_table["*"])
			t = t:gsub("%_", escape_table["_"])
		else
			t = encode_backslash_escapes(t)
		end
		out = out .. t
	end
	return out
end

-- Encode backspace-escaped characters in the markdown source.
function encode_backslash_escapes(t)
	for i=1,escape_chars:len() do
		local c = escape_chars:sub(i,i)
		t = t:gsub("\\%" .. c, escape_table[c])
	end
	return t
end

-- Unescape characters that have been encoded.
function unescape_special_chars(t)
	for k,v in pairs(escape_table) do
		t = t:gsub(v,k)
	end
	return t
end

-- Encode/escape certain characters inside Markdown code runs.
-- The point is that in code, these characters are literals,
-- and lose their special Markdown meanings.
function encode_code(s)
	s = s:gsub("%&", "&amp;")
	s = s:gsub("<", "&lt;")
	s = s:gsub(">", "&gt;")
	for k,v in pairs(escape_table) do
		s = s:gsub("%"..k, v)
	end
	return s
end

-- Handle backtick blocks.
function code_spans(s)
	s = s:gsub("\\\\", escape_table["\\"])
	s = s:gsub("\\`", escape_table["`"])

	local pos = 1
	while true do
		local start, stop = s:find("`+", pos)
		if not start then return s end
		local count = stop - start + 1
		-- Find a matching numbert of backticks
		local estart, estop = s:find(string.rep("`", count), stop+1)
		local brstart = s:find("\n", stop+1)
		if estart and (not brstart or estart < brstart) then
			local code = s:sub(stop+1, estart-1)
			code = code:gsub("^[ \t]+", "")
			code = code:gsub("[ \t]+$", "")
			code = code:gsub(escape_table["\\"], escape_table["\\"] .. escape_table["\\"])
			code = code:gsub(escape_table["`"], escape_table["\\"] .. escape_table["`"])
			code = "<code>" .. encode_code(code) .. "</code>"
			s = s:sub(1, start-1) .. code .. s:sub(estop+1)
			pos = start + code:len()
		else
			pos = stop + 1
		end
	end
	return s
end

-- Handle image references
function images(text)
	local function reference_link(alt, id)
		alt = alt:match("%[(.*)%]")
		id = id:match("%[(.*)%]"):lower()
		if id == "" then id = text:lower() end
		link_database[id] = link_database[id] or {}
		if not link_database[id].url then return nil end
		local url = link_database[id].url or id
		local title = link_database[id].title
		if title then title = " title=\"" .. title .. "\"" else title = "" end
		return '<img src="' .. url .. '" alt="' .. alt .. '"' .. title .. "/>"
	end
	
	local function inline_link(alt, link)
		alt = alt:match("%[(.-)%]")
		local url, title = link:match("%(<?(.-)>?[ \t]*['\"](.+)['\"]")
		url = url or link:match("%(<?(.-)>?%)")
		if title then
			title = title:gsub('"', '&quot;')
			return '<img src="' .. url .. '" alt="' .. alt .. '" title="' .. title .. '"/>'
		else
			return '<img src="' .. url .. '" alt="' .. alt .. '"/>'
		end
	end
	
	text = text:gsub("!(%b[])[ \t]*\n?[ \t]*(%b[])", reference_link)
	text = text:gsub("!(%b[])[ \t]*\n?[ \t]*(%b())", inline_link)
	return text
end

-- Handle anchor references
function anchors(text)
	local function reference_link(text, id)
		text = text:match("%[(.*)%]")
		id = id:match("%[(.*)%]"):lower()
		if id == "" then id = text:lower() end
		link_database[id] = link_database[id] or {}
		if not link_database[id].url then return nil end
		local url = link_database[id].url or id
		local title = link_database[id].title
		if title then title = " title=\"" .. title .. "\"" else title = "" end
		return "<a href=\"" .. url .. "\"" .. title .. ">" .. text .. "</a>"
	end
	
	local function inline_link(text, link)
		text = text:match("%[(.-)%]")
		local url, title = link:match("%(<?(.-)>?[ \t]*['\"](.+)['\"]")
		url  = url or  link:match("%(<?(.-)>?%)") or ""
		if title then
			title = title:gsub('"', '&quot;')
			return "<a href=\"" .. url .. "\" title=\"" .. title .. "\">" .. text .. "</a>"
		else
			return "<a href=\"" .. url .. "\">" .. text .. "</a>"
		end
	end
	
	text = text:gsub("(%b[])[ \t]*\n?[ \t]*(%b[])", reference_link)
	text = text:gsub("(%b[])[ \t]*\n?[ \t]*(%b())", inline_link)
	return text
end

-- Handle auto links, i.e. <http://www.google.com/>.
function auto_links(text)
	local function link(s)
		return "<a href=\"" .. s .. "\">" .. s .. "</a>"
	end
	-- Encode chars as a mix of dec and hex entitites to (perhaps) fool
	-- spambots.
	local function encode_email_address(s)
		-- Use a deterministic encoding to make unit testing possible.
		-- Code 45% hex, 45% dec, 10% plain.
		local hex = {code = function(c) return "&#x" .. string.format("%x", c:byte()) .. ";" end, count = 1, rate = 0.45}
		local dec = {code = function(c) return "&#" .. c:byte() .. ";" end, count = 0, rate = 0.45}
		local plain = {code = function(c) return c end, count = 0, rate = 0.1}
		local codes = {hex, dec, plain}
		local function swap(t,k1,k2) local temp = t[k2] t[k2] = t[k1] t[k1] = temp end
		
		local out = ""
		for i = 1,s:len() do
			for _,code in ipairs(codes) do code.count = code.count + code.rate end
			if codes[1].count < codes[2].count then swap(codes,1,2) end
			if codes[2].count < codes[3].count then swap(codes,2,3) end
			if codes[1].count < codes[2].count then swap(codes,1,2) end
			
			local code = codes[1]
			local c = s:sub(i,i)
			-- Force encoding of "@" to make email address more invisible.
			if c == "@" and code == plain then code = codes[2] end
			out = out .. code.code(c)
			code.count = code.count - 1
		end	
		return out
	end
	local function mail(s)
		s = unescape_special_chars(s)
		local address = encode_email_address("mailto:" .. s)
		local text = encode_email_address(s)
		return "<a href=\"" .. address .. "\">" .. text .. "</a>"
	end
	-- links
	text = text:gsub("<(https?:[^'\">%s]+)>", link)
	text = text:gsub("<(ftp:[^'\">%s]+)>", link)
	
	-- mail
	text = text:gsub("<mailto:([^'\">%s]+)>", mail)
	text = text:gsub("<([-.%w]+%@[-.%w]+)>", mail)
	return text
end

-- Encode free standing amps (&) and angles (<)... note that this does not
-- encode free >.
function amps_and_angles(s)
	-- encode amps not part of &..; expression
	local pos = 1
	while true do
		local amp = s:find("&", pos)
		if not amp then break end
		local semi = s:find(";", amp+1)
		local stop = s:find("[ \t\n&]", amp+1)
		if not semi or (stop and stop < semi) or (semi - amp) > 15 then
			s = s:sub(1,amp-1) .. "&amp;" .. s:sub(amp+1)
			pos = amp+1
		else
			pos = amp+1
		end
	end
	
	-- encode naked <'s
	s = s:gsub("<([^a-zA-Z/?$!])", "&lt;%1")
	
	-- what about >, nothing done in the original markdown source to handle them
	return s
end

-- Handles emphasis markers (* and _) in the text.
function emphasis(text)
	for _, s in ipairs {"%*%*", "%_%_"} do
		text = text:gsub(s .. "([^%s][%*%_]?)" .. s, "<strong>%1</strong>")
		text = text:gsub(s .. "([^%s][^<>]-[^%s][%*%_]?)" .. s, "<strong>%1</strong>")
	end
	for _, s in ipairs {"%*", "%_"} do
		text = text:gsub(s .. "([^%s])" .. s, "<em>%1</em>")
		text = text:gsub(s .. "([^%s][^<>]-[^%s])" .. s, "<em>%1</em>")
	end
	return text
end

-- Handles line break markers in the text.
function line_breaks(text)
	return text:gsub("  +\n", " <br/>\n")
end

-- Perform all span level transforms.
function span_transform(text)
	text = code_spans(text)
	text = escape_special_chars(text)
	text = images(text)
	text = anchors(text)
	text = auto_links(text)
	text = amps_and_angles(text)
	text = emphasis(text)
	text = line_breaks(text)
	return text
end

----------------------------------------------------------------------
-- Markdown
----------------------------------------------------------------------

-- Cleanup the text by normalizing some possible variations to make further
-- processing easier.
function cleanup(text)
	-- Standardize line endings
	text = text:gsub("\r\n", "\n")  -- DOS to UNIX
	text = text:gsub("\r", "\n")    -- Mac to UNIX
	
	-- Convert all tabs to spaces
	text = detab(text)
	
	-- Strip lines with only spaces and tabs
	while true do
		local subs
		text, subs = text:gsub("\n[ \t]+\n", "\n\n")
		if subs == 0 then break end
	end
	
	return "\n" .. text .. "\n"
end

-- Strips link definitions from the text and stores the data in a lookup table.
function strip_link_definitions(text)
	local linkdb = {}
	
	local function link_def(id, url, title)
		id = id:match("%[(.+)%]"):lower()
		if title then title = title:gsub('"', '&quot;') end
		linkdb[id] = {url = url, title = title}
		return ""
	end

	local def_no_title = "\n ? ? ?(%b[]):[ \t]*\n?[ \t]*<?([^%s>]+)>?"
	local def_title = def_no_title .. "[ \t]*\n?[ \t]*[\"'(]([^\n]+)[\"')]"
	
	text = text:gsub(def_title, link_def)
	text = text:gsub(def_no_title, link_def)
	return text, linkdb
end

link_database = {}

-- Main markdown processing function
function markdown(text)
	init_hash(text)
	init_escape_table()
	
	text = cleanup(text)
	text = protect(text)
	text, link_database = strip_link_definitions(text)
	text = block_transform(text)
	text = unescape_special_chars(text)
	return text
end

----------------------------------------------------------------------
-- End of module
----------------------------------------------------------------------

setfenv(1, _G)
M.lock(M)

-- Expose markdown function to the world
markdown = M.markdown

-- Class for parsing command-line options
local OptionParser = {}
OptionParser.__index = OptionParser

-- Creates a new option parser
function OptionParser:new()
	local o = {short = {}, long = {}}
	setmetatable(o, self)
	return o
end

-- Calls f() whenever a flag with specified short and long name is encountered
function OptionParser:flag(short, long, f)
	local info = {type = "flag", f = f}
	if short then self.short[short] = info end
	if long then self.long[long] = info end
end

-- Calls f(param) whenever a parameter flag with specified short and long name is encountered
function OptionParser:param(short, long, f)
	local info = {type = "param", f = f}
	if short then self.short[short] = info end
	if long then self.long[long] = info end
end

-- Calls f(v) for each non-flag argument
function OptionParser:arg(f)
	self.arg = f
end

-- Runs the option parser for the specified set of arguments. Returns true if all arguments
-- where successfully parsed and false otherwise.
function OptionParser:run(args)
	local pos = 1
	while pos <= #args do
		local arg = args[pos]
		if arg == "--" then
			for i=pos+1,#args do
				if self.arg then self.arg(args[i]) end
				return true
			end
		end
		if arg:match("^%-%-") then
			local info = self.long[arg:sub(3)]
			if not info then print("Unknown flag: " .. arg) return false end
			if info.type == "flag" then
				info.f()
				pos = pos + 1
			else
				param = args[pos+1]
				if not param then print("No parameter for flag: " .. arg) return false end
				info.f(param)
				pos = pos+2
			end
		elseif arg:match("^%-") then
			for i=2,arg:len() do
				local c = arg:sub(i,i)
				local info = self.short[c]
				if not info then print("Unknown flag: -" .. c) return false end
				if info.type == "flag" then
					info.f()
				else
					if i == arg:len() then
						param = args[pos+1]
						if not param then print("No parameter for flag: -" .. c) return false end
						info.f(param)
						pos = pos + 1
					else
						param = arg:sub(i+1)
						info.f(param)
					end
					break
				end
			end
			pos = pos + 1
		else
			if self.arg then self.arg(arg) end
			pos = pos + 1
		end
	end
	return true
end

-- Handles the case when markdown is run from the command line
local function run_command_line(arg)
	-- Generate output for input s given options
	local function run(s, options)
		s = markdown(s)
		if not options.wrap_header then return s end
		local header = ""
		if options.header then
			local f = io.open(options.header) or error("Could not open file: " .. options.header) 
			header = f:read("*a")
			f:close()
		else
			header = [[
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html>
<head>
	<meta http-equiv="content-type" content="text/html; charset=CHARSET" />
	<title>TITLE</title>
	<link rel="stylesheet" type="text/css" href="STYLESHEET" />
</head>
<body>
]]
			local title = options.title or s:match("<h1>(.-)</h1>") or s:match("<h2>(.-)</h2>") or 
				s:match("<h3>(.-)</h3>") or "Untitled"
			header = header:gsub("TITLE", title)
			if options.inline_style then
				local style = ""
				local f = io.open(options.stylesheet)
				if f then 
					style = f:read("*a") f:close() 
				else
					error("Could not include style sheet " .. options.stylesheet .. ": File not found")
				end
				header = header:gsub('<link rel="stylesheet" type="text/css" href="STYLESHEET" />',
					"<style type=\"text/css\"><!--\n" .. style .. "\n--></style>")
			else
				header = header:gsub("STYLESHEET", options.stylesheet)
			end
			header = header:gsub("CHARSET", options.charset)
		end
		local footer = "</body></html>"
		if options.footer then
			local f = io.open(options.footer) or error("Could not open file: " .. options.footer)
			footer = f:read("*a")
			f:close()
		end
		return header .. s .. footer
	end
	
	-- Generate output path name from input path name given options.	
	local function outpath(path, options)
		if options.append then return path .. ".html" end
		local m = path:match("^(.+%.html)[^/\\]+$") if m then return m end
		m = path:match("^(.+%.)[^/\\]*$") if m and path ~= m .. "html" then return m .. "html" end
		return path .. ".html"
	end
	
	-- Default commandline options
	local options = {
		wrap_header = true,
		header = nil,
		footer = nil,
		charset = "utf-8",
		title = nil,
		stylesheet = "default.css",
		inline_style = false
	}
	local help = [[
Usage: markdown.lua [OPTION] [FILE]
Runs the markdown text markup to HTML converter on each file specified on the
command line. If no files are specified, runs on standard input.

No header:
    -n, --no-wrap        Don't wrap the output in <html>... tags.
Custom header:
    -e, --header FILE    Use content of FILE for header.
    -f, --footer FILE    Use content of FILE for footer.
Generated header:
    -c, --charset SET    Specifies charset (default utf-8).
    -i, --title TITLE    Specifies title (default from first <h1> tag).
    -s, --style STYLE    Specifies style sheet file (default default.css).
	-l, --inline-style   Include the style sheet file inline in the header.
Generated files:
    -a, --append         Append .html extension (instead of replacing).
Other options:
    -h, --help           Print this help text.
    -t, --test           Run the unit tests.
]]

	local run_stdin = true
	local op = OptionParser:new()
	op:flag("n", "no-wrap", function () options.wrap_header = false end)
	op:param("e", "header", function (x) options.header = x end)
	op:param("f", "footer", function (x) options.footer = x end)
	op:param("c", "charset", function (x) options.charset = x end)
	op:param("i", "title", function(x) options.title = x end)
	op:param("s", "style", function(x) options.stylesheet = x end)
	op:flag("l", "inline-style", function(x) options.inline_style = true end)
	op:flag("a", "append", function() options.append = true end)
	op:flag("t", "test", function() 
		local n = arg[0]:gsub("markdown.lua", "markdown-tests.lua")
		local f = io.open(n)
		if f then 
			f:close() dofile(n) 
		else
			error("Cannot find markdown-tests.lua")
		end
		run_stdin = false 
	end)
	op:flag("h", "help", function() print(help) run_stdin = false end)
	op:arg(function(path) 
			local file = io.open(path) or error("Could not open file: " .. path)
			local s = file:read("*a")
			file:close()
			s = run(s, options)
			file = io.open(outpath(path, options), "w") or error("Could not open output file: " .. outpath(path, options))
			file:write(s)
			file:close()
			run_stdin = false
		end
	)
	
	if not op:run(arg) then
		print(help)
		run_stdin = false
	end

	if run_stdin then
		local s = io.read("*a")
		s = run(s, options)
		io.write(s)
	end
end
	
-- If we are being run from the command-line, act accordingly
if arg[0]:find("markdown%.lua$") then
	run_command_line(arg)
else
	return markdown
end