How do I convert a Latex document containing Tikz pictures to HTML?

[minoush82]

I need to convert a Latex document containing Tikz pictures to HTML. The following is an example document example.tex:

\documentclass[a4paper]{report}
\usepackage{tikz}
\usetikzlibrary{automata, positioning, arrows}
\begin{document}
\tikzset{->, >=stealth', node distance=3cm, every state/.style={thick, fill=gray!10}, initial text=$ $,}
The following is a TIKZ picture.
\begin{tikzpicture}
	\node[state, initial, accepting] (q1) {$q_1$};
	\node[state, right of=q1] (q2) {$q_2$};
	\node[state, right of=q2] (q3) {$q_3$};
	\node[state, accepting, right of=q3] (q4) {$q_4$};
	\draw (q1) edge[loop above] node{a} (q1)
	(q1) edge[above] node{b} (q2)
	(q2) edge[above] node{a} (q3)
	(q2) edge[below, bend right] node{b} (q4)
	(q3) edge[above, bend right] node{a} (q2)
	(q3) edge[above, bend left] node{b} (q1)
	(q4) edge[loop above] node{b} (q4)
	(q4) edge[above] node{a} (q3);
\end{tikzpicture}
The above is a TIKZ picture.
\end{document}

which has the following PDF output.

In order to convert to HTML, I am using the following command:

pandoc --standalone -N --lua-filter=diagram.lua -f latex -t html example.tex -o example.html

where diagram.lua is this one: https://github.com/pandoc-ext/diagram/blob/main/_extensions/diagram/diagram.lua

Unfortunately, the HTML output I get is the following where the picture is missing.

The following is a TIKZ picture.

The above is a TIKZ picture.

Which is the same output I get when I don't use the filter in the pandoc command.

Could you please help me achieve my goal?

[jgm]

Here's what I do in a book I'm working on. I use a Lua filter tex2image.lua:

local system = require 'pandoc.system'

local doc_template = [[
\documentclass{standalone}
\usepackage{bookgraphics} % Local package with my own macro definitions used in tikz diagrams
\begin{document}
%s
\end{document}
]]

local function tex2image(src, filetype, outfile)
    local f = io.open('tex-temp.tex', 'w')
    f:write(doc_template:format(src))
    f:close()
    local _, _, status = os.execute('lualatex -pdflua -auxdir=textmp -interaction=nonstopmode -halt-on-error -shell-escape tex-temp.tex')
    if status ~= 0 then
      os.exit(1)
    end
    if filetype == 'pdf' then
        os.rename('tex-temp.pdf', outfile)
     else
        os.execute('pdf2svg tex-temp.pdf ' .. outfile)
    end
end

extension_for = {
  html = 'svg',
  html4 = 'svg',
  html5 = 'svg',
  latex = 'pdf',
  beamer = 'pdf' }

local function file_exists(name)
  local f = io.open(name, 'r')
  if f ~= nil then
    io.close(f)
    return true
  else
    return false
  end
end

function RawBlock(el)
  if el.format == 'latex' and
    (el.text:match("^\\begin%{tikz") or el.text:match("^%$")) then
    local filetype = extension_for[FORMAT] or 'svg'
    local fbasename = pandoc.sha1(el.text) .. '.' .. filetype
    if not pandoc.mediabag.lookup(fbasename) then
      local outname = "tex-temp." .. filetype
      tex2image(el.text, filetype, outname)
      local mt, contents = pandoc.mediabag.fetch(outname)
      pandoc.mediabag.insert(fbasename, mt, contents)
    end
    return pandoc.Para({pandoc.Image({}, fbasename)})
  else
   return el
  end
end

This uses lualatex and pdf2svg to generate svg versions of the tikz environments, which it then puts in the "mediabag." My source format is markdown, but I think it would work for you if you use -f latex+raw_tex. You will also want --extract-media=media which will create a media subdirectory with the svgs linked from the HTML file.

diagram.lua is, if I recall, designed to translate certain markdown code blocks into tikz diagrams, so it's not going to work for going directly from latex.

[jgm]

Note: where I have \usepackage{bookgraphics} in the template, you'll need to put whatever you need to compile your tikz, e.g.

\usepackage{tikz}
\usetikzlibrary{automata, positioning, arrows}

[minoush82]

Many thanks :-) Your solution works :-)

The following input file (main.tex):

\documentclass[a4paper]{report}
\usepackage{tikz}
\usetikzlibrary{automata, positioning, arrows}
\tikzset{->, >=stealth', node distance=3cm, every state/.style={thick, fill=gray!10}, initial text=$ $,}
\begin{document}
The following is a TIKZ picture.
\begin{tikzpicture}
	\node[state, initial, accepting] (q1) {$q_1$};
	\node[state, right of=q1] (q2) {$q_2$};
	\node[state, right of=q2] (q3) {$q_3$};
	\node[state, accepting, right of=q3] (q4) {$q_4$};
	\draw (q1) edge[loop above] node{a} (q1)
	(q1) edge[above] node{b} (q2)
	(q2) edge[above] node{a} (q3)
	(q2) edge[below, bend right] node{b} (q4)
	(q3) edge[above, bend right] node{a} (q2)
	(q3) edge[above, bend left] node{b} (q1)
	(q4) edge[loop above] node{b} (q4)
	(q4) edge[above] node{a} (q3);
\end{tikzpicture}
The above is a TIKZ picture.
\end{document}

has successfully been converted to HTML with SVG figures using the command:
pandoc --standalone -N --lua-filter=tex2image.lua --extract-media=media -f latex+raw_tex -t html main.tex -o main.html

where tex2image.lua is the following:

local system = require 'pandoc.system'

local doc_template = [[
\documentclass[a4paper]{report}
\usepackage{tikz}
\usetikzlibrary{automata, positioning, arrows}
\tikzset{->, >=stealth', node distance=3cm, every state/.style={thick, fill=gray!10}, initial text=$ $,}
\begin{document}
%s
\end{document}
]]

local function tex2image(src, filetype, outfile)
    local f = io.open('tex-temp.tex', 'w')
    f:write(doc_template:format(src))
    f:close()
    local _, _, status = os.execute('lualatex -pdflua -auxdir=textmp -interaction=nonstopmode -halt-on-error -shell-escape tex-temp.tex')
    if status ~= 0 then
      os.exit(1)
    end
    if filetype == 'pdf' then
        os.rename('tex-temp.pdf', outfile)
     else
        os.execute('pdf2svg tex-temp.pdf ' .. outfile)
    end
end

extension_for = {
  html = 'svg',
  html4 = 'svg',
  html5 = 'svg',
  latex = 'pdf',
  beamer = 'pdf' }

local function file_exists(name)
  local f = io.open(name, 'r')
  if f ~= nil then
    io.close(f)
    return true
  else
    return false
  end
end

function RawBlock(el)
  if el.format == 'latex' and
    (el.text:match("^\\begin%{tikz") or el.text:match("^%$")) then
    local filetype = extension_for[FORMAT] or 'svg'
    local fbasename = pandoc.sha1(el.text) .. '.' .. filetype
    if not pandoc.mediabag.lookup(fbasename) then
      local outname = "tex-temp." .. filetype
      tex2image(el.text, filetype, outname)
      local mt, contents = pandoc.mediabag.fetch(outname)
      pandoc.mediabag.insert(fbasename, mt, contents)
    end
    return pandoc.Para({pandoc.Image({}, fbasename)})
  else
   return el
  end
end

The obtained output is the following:

My only remaining issue is that the bounding box of the resulting pictures is quite large as can be seen in the following screenshot from Inkscape.

Any ideas on how I might solve this issue?

[iandol]

See my general comment, dvisvgm claims to tighten bounding boxes:

Computes tight bounding boxes for the generated graphics

[jgm]

I think the issue is that you are using the report document class. Use standalone as in my original filter.

[minoush82]

@jgm, you are absolutely right, I completely overlooked the standalone class. Now it works fine.

[iandol]

There is another Lua filter you could try: https://dialoa.github.io/imagify/ -- it uses a different SVG converter (https://dvisvgm.de rather than PDF2SVG) IINM. You could also edit John's filter to use dvisvgm?

[minoush82]

Thank you, I shall look into this.