.latex_to_html5

#!/usr/bin/env lua5.3

-- lua filter for converting (a subset of) LaTeX into HTML5 with mathjax
-- Basic idea: leave the math to matjax and search&replace simple commands
-- Note that this does not really parse the LaTeX source, it only performs
-- a line-based pattern matching for a well-behaved subset of LaTeX that
-- i am forced to use if i want to run this filter.

-- read whole input file into a string
whole_file = io.read("*all")

-- extract title and other data
metadata_title = whole_file:match("^\\title{(.-)}")
metadata_maths = false

-- table of global substitutions (could be implemented in sed)
global_substitutions = {

	-- change a few character entities
	{"&", "&"},   -- note: this affects LaTeX tables, see below
	{"<", "&lt;"},
	{">", "&gt;"},
	{"~", " "}, -- real non-breaking space here, equivalent to &nbsp;

	-- htmlize SCRIPT tags
	{"\n%%(SCRIPT[^\n]+)", "\n<!--\n&#37;%1\n-->"},
	{"\n%-%->\n<!%-%-", ""},

	-- do not remove comments, just escape % to &#37;
	{"%%", "&#37;"},

	-- change \emph{...} to <i>...</i> (and similar inline stuff)
	{"\\emph{(.-)}", "<i>%1</i>"},
	{"\\textbf{(.-)}", "<b>%1</b>"},
	{"\\textit{(.-)}", "<i>%1</i>"},
	{"\\texttt{(.-)}", "<code>%1</code>"},
	{"{\\it%s(.-)}", "<i>%1</i>"},
	{"{\\bf%s(.-)}", "<b>%1</b>"},
	{"{\\tt%s(.-)}", "<code>%1</code>"},

	-- change \includegraphics{...} to <img src="...">
	{"\\includegraphics{(.-)}", '<img src="%1" alt="image">'},

	-- change \href{link}{text} to <a href="link">text</a>
	{"\\href{(.-)}{(.-)}", "<a href=\"%1\">%2</a>"},
	{"\\url{(.-)}", "<a href=\"%1\">%1</a>"},

	-- change \newline to <br>
	{"\\newline", "<br>"},
	{"\\bigskip", "<p>"},
	{"\\medskip", "<p>"},
	{"\\smallskip", ""},

	-- remove latex-only stuff
	{"\\clearpage", ""},
	{"\\thispagestyle{(.-)}", ""},
	{"\\setcounter{(.-)}{(.-)}", ""},
	{"\\vfill", ""},
	{"\\mbox{(.-)}", '<span>%1</span>'},
	{"\\fbox{(.-)}", '<span style="border:1px solid">%1</span>'},

	-- change \sections to <headers>
	{"\\title{(.-)}", "<h1>%1</h1>\n"},
	{"\\section{(.-)}", "<h2>%1</h2>\n"},
	{"\\subsection{(.-)}", "<h3>%1</h3>\n"},
	{"\\paragraph{(.-)}", "<p>\n<b>%1</b>"},

	-- list environments (TODO: support for \item[] )
	{"\\item ", "<li>" },
	{"\\begin{enumerate}", "<ol>"},
	{"\\end{enumerate}", "</ol>"},
	{"\\begin{itemize}", "<ul>"},
	{"\\end{itemize}", "</ul>"},
	{"\\begin{verbatim}", "<pre>"},
	{"\\end{verbatim}", "</pre>"},
	{"\\item%[(.-)%] ", "<li style='list-style-type:none'>%1 "},

	-- change \verb+...+ to <code>...</code>
	{"\\verb%+([^+]*)%+", "<code>%1</code>"},
	{"\\verb_([^_]*)_", "<code>%1</code>"},

	-- quotation environements
	{"\\begin{quotation}", "<blockquote>"},
	{"\\end{quotation}", "</blockquote>"},
	{"\\begin{quote}", "<blockquote>"},
	{"\\end{quote}", "</blockquote>"},

	-- gallery environement
	{"\\begin{gallery}", "<div class=\"gallery\" style=\"height:400px\"><ul class=\"index\">\n"},
	{"\\end{gallery}", "</ul></div>"},
	{"\\galleryline{(.-)}", '<li><a href="%1">%1<span><img src="%1" alt=""></span></a>\n'},

	-- theorem environments (TODO: different class for each type)
	{"\\begin{proof}",      "<p>\n<i>Proof.</i>"},
	{"\\end{proof}",        "<span style=\"float:right\">&#8718;</span>"},
	-- to avoid repetitive code, the other environments are added in a loop
	-- below

	-- abstract environement (wherever it appears)
	{"\\begin{abstract}(.-)\\end{abstract}",
	 '<blockquote style="font-size:80%%"><b>Abstract.</b>%1</blockquote>'
	},

	-- tabular (TODO: recover vertical bars)
	-- XXX: the order of these pattern replacements is important
	{"%s*\\\\(%s*\\end{tabular})", "%1"},
	{"\\begin{tabular}{.-}", "<table><tr><td>"},
	{"\\end{tabular}", "</table>"},

	-- surround \newcommands by math so that mathjax sees it
	{"\n(\\newcommand[^\n]+)", "$%1$\n"},

	-- hack to remove "i/" from displayed image and text filenames
	{'([^"])i/([%a.-]-%.tif)', "%1%2"},
	{'([^"])i/([%a.-]-%.png)', "%1%2"},
	{'([^"])i/([%a.-]-%.jpg)', "%1%2"},
	{'([^"])i/([%a.-]-%.txt)', "%1%2"},
	{'([^"])o/([%a.-]-%.tif)', "%1%2"},
	{'([^"])o/([%a.-]-%.png)', "%1%2"},
	{'([^"])o/([%a.-]-%.jpg)', "%1%2"},
	{'([^"])o/([%a.-]-%.txt)', "%1%2"},
}

-- add theorem-like environments to the list of substitutions
thm_envs = {"Theorem", "Definition", "Remark", "Lemma", "Proposition",
	"Exercice", "Axiom"}
for _,X in pairs(thm_envs) do
	x = X:lower()
	p = {"\\begin{"..x.."}%[(.-)%]",
		'<p>\n<div class="thm"><b>'..X..'</b> (%1). <em>'}
	q = {"\\begin{"..x.."}", '<p>\n<div class="thm"><b>'..X.."</b>. <em>"}
	r = {"\\end{"..x.."}", "</em></div>"}
	table.insert(global_substitutions, p)
	table.insert(global_substitutions, q)
	table.insert(global_substitutions, r)
end

-- perform the global substitutions
for _,s in ipairs(global_substitutions) do
	whole_file = whole_file:gsub(s[1], s[2])
end

-- process the galleries
list_of_galleries = {}

-- g traverses all gallery blocks
for g in whole_file:gmatch('(<div class="gallery" .-</ul></div>)') do
	f = g:match('<img src="(.-)"') -- f=first image file in the gallery
	h = io.popen("imprintf %h "..f):read("*all*") -- h=image height
	_,n = g:gsub("<li>", "") -- n=total number of images in the gallery
	table.insert(list_of_galleries, {f, h, math.ceil(2*n)})
end

-- adapt the size of each gallery
for i=1,#list_of_galleries do
	g = list_of_galleries[i]
	f,h,n = g[1],g[2],g[3]    -- filename, height in pixels, height in em
	f2 = f:gsub("%-", "--")   -- escaped filename
	p = 'gallery" style="height:400px">(.-<a href=")'..f2
	q = 'gallery" style="min-height:'..h..'px;height:'..n..'em">%1'..f
	whole_file = whole_file:gsub(p, q)
end

-- process \input{file} and \VerbatimInput{file}
putfile = function(l) return io.open(l, "r"):read("*all") end
putfile_verb = function(l) return "<pre>\n"..io.open(l, "r"):read("*all").."</pre>\n" end
whole_file = whole_file:gsub("\\input{(.-)}", putfile)
whole_file = whole_file:gsub("\\VerbatimInput{(.-)}", putfile_verb)


output_template = [[
<!doctype html>
<meta charset="utf-8" />
<title>TITLE</title>
CSSCODE
JAXCODE
LITCODE
]]

boilerplate_lit = [[
<!--
	This file is a literate program.
	The experiments are run by applying the following filter:

		grep ^%%SCRIPT | sed 's/&amp;/\&/g' | cut -c9- | sh
-->
]]

boilerplate_css = [[
	body { max-width:90ex; }
	pre { background:lightgray; width:80ch; }
	table, td { border:1px solid black; border-collapse:collapse; }
	table td { padding:7px; border-spacing:0px; }
	.thm em i { font-style: normal;}]]
-- TODO: only add the styling if the concerned elements are present

css_gallery = [[
	.gallery{position:relative;width:auto;height:400px}
	.gallery .index{padding:0;margin:0;width:9em;list-style:none}
	.gallery .index li{margin:0;padding:0}
	.gallery .index a{display:block;background-color:#eee;border:1px solid #fff;text-decoration:none;width:11em;padding:5px}
	.gallery .index a span{display:block;position:absolute;left:-9999px;top:0;padding-left:2em}
	.gallery .index li:first-child a span{left:10em;z-index:99}
	.gallery .index a:hover{ border: 1px solid #888888;}
	.gallery .index a:hover span{left:10em;z-index:100}
	.gallery .index a span img{ }
	.gallery .index a span { white-space:nowrap; }
]]

boilerplate_mathjax = [[
<script type="text/x-mathjax-config">
	MathJax.Hub.Config({
		tex2jax: {
			inlineMath: [ ['$','$'], ['\\(','\\)'] ],
			processEscapes: true
		}
	});
</script>
<script async src="JAXURL?JAXOPT">
</script>
]]

JAXURL = "https://cdn.rawgit.com/mathjax/MathJax/2.7.1/MathJax.js"
JAXOPT = "config=TeX-AMS_CHTML-full"

if whole_file:match("%$") then  -- TODO: perform a more intelligent check here
	metadata_maths = true
end
if not metadata_maths then boilerplate_mathjax = "" end

if not whole_file:match("\n&#37;SCRIPT") then boilerplate_lit = "" end

if whole_file:match("<div class=\"gallery\"") then
	boilerplate_css = boilerplate_css .. css_gallery
end

boilerplate_mathjax = boilerplate_mathjax:gsub("JAXURL", JAXURL)
boilerplate_mathjax = boilerplate_mathjax:gsub("JAXOPT", JAXOPT)

if not metadata_title then metadata_title = "" end

boilerplate_css = "<style>\n"..boilerplate_css.."\n</style>"
output_template = output_template:gsub("TITLE", metadata_title)
output_template = output_template:gsub("CSSCODE", boilerplate_css)
output_template = output_template:gsub("JAXCODE", boilerplate_mathjax)
output_template = output_template:gsub("LITCODE", boilerplate_lit)

-- start building output file
output_lines = { output_template }
table.insert(output_lines, "\n")


-- initialize counters
counter_sec = 0
counter_ssec = 0
counter_thm = 0
table_depth = 0
toggle_pre = false

o = "" -- the previous line
for l in string.gmatch(whole_file, "[^\n]-\n") do -- traverse line by line
	u = l    -- line to add to the output (the same, by default)

	-- if a paragraph starts, put <p>
	if o == "\n" and
		not toggle_pre and
		not u:match("^%$\\newcommand") and
		(
			u:match("^[%a][^\n]-\n") or
			u:match("^<img[^\n]-\n") or
			u:match("^<span[^\n]-\n") or
			u:match("^<b>[^\n]-\n") or
			u:match("^<i>[^\n]-\n") or
			u:match("^<em>[^\n]-\n") or
			u:match("^%$[^%$]")
		)
		then
		u = "<p>\n"..u
	end

	-- build a table context
	if o:match("<table"  ) then table_depth = table_depth + 1 end
	if u:match("</table>") then table_depth = table_depth - 1 end

	-- if inside table, change & and \\ to <td> and <tr>, remove \hline
	if table_depth > 0 then
		u = u:gsub("&amp;", "<td>")
		u = u:gsub("\\\\", "<tr><td>")
		u = u:gsub("\\hline",  "")
	end

	-- section counters (TODO: add attr name="sec-n" for easier navigation)
	if u:match("^<h2>.*$") then
		counter_sec = counter_sec + 1
		counter_ssec = 0
		u = u:gsub("^<h2>", "<h2>"..counter_sec..". ")
	end

	-- subsection counters
	if u:match("^<h3>.*$") then
		counter_ssec = counter_ssec + 1
		u = u:gsub("^<h3>","<h3>"..counter_sec.."."..counter_ssec..". ")
	end

	-- theorem counters
	if u:match("^<div class=\"thm\">.*%</b>") then
		counter_thm = counter_thm + 1
		u = u:gsub("%</b>", " "..counter_thm.."</b>")
	end

	-- <pre> book-keeping
	if u:match("^<pre") then toggle_pre = true end
	if u:match("^ <pre") then toggle_pre = true end
	if u:match("</pre>") then toggle_pre = false end

	-- post-processing replacements
	if toggle_pre then                -- if inside <pre> :
		u = u:gsub("&#37;", "%%") -- un-escape html entity for %
	end
	if u:match("^&#37;%SCRIPT ") then -- if inside valid ^%SCRIPT line :
		u = u:gsub("&lt;", "<")   -- un-escape html entity for <
		u = u:gsub("&gt;", ">")   -- un-escape html entity for >
		u = u:gsub("&#37;", "%%") -- un-escape html entity for %
	elseif not toggle_pre then        -- deal with comments
		u = u:gsub("^&#37[^\n]*\n", "") -- remove whole line comment
		u = u:gsub("([^\\])&#37[^\n]*\n", "%1\n") -- midline comment
		u = u:gsub("\\&#37;", "\\%%") -- escaped % symbols
	end

	if o ~= "\n" or u ~= "\n" then -- uniq-ify blank lines
		table.insert(output_lines, u)
	end
	o = l
end

table.insert(output_lines, "\n<hr>\n\n")
--table.insert(output_lines, '\n<!-- yes i know, wanna fight about it ? -->\n')
--tracker = string.format('<img src="%s?a=%s" width="1" height="1" alt="">\n\n',
--	"http://boucantrin.ovh.hw.ipol.im:7743/white_pixel.png",
--	metadata_title:gsub("%s", "_")
--	)
--table.insert(output_lines, tracker)
table.insert(output_lines, "<!-- enric meinhardt-llopis, 2019 -->\n\n")

-- dump output lines
for _,l in ipairs(output_lines) do
	io.stdout:write(l)
end