First attempt at generating structure elements

luamml-convert.lua

@@ -129,7 +129,7 @@ local function delim_to_table(delim)
   else
     local fam = delim.small_fam
     char = remap_lookup[fam << 21 | char]
-    local result = {[0] = 'mo', char, ['tex:family'] = fam ~= 0 and fam or nil, stretchy = not stretchy[char] or nil, lspace = 0, rspace = 0 }
+    local result = {[0] = 'mo', char, ['tex:family'] = fam ~= 0 and fam or nil, stretchy = not stretchy[char] or nil, lspace = 0, rspace = 0, [':node'] = delim }
     return result, result
   end
 end
@@ -149,7 +149,7 @@ local function acc_to_table(acc, cur_style, stretch)
   if stretch ~= not stretchy[char] then -- Handle nil gracefully in stretchy
     stretch = nil
   end
-  return {[0] = 'mo', char, ['tex:family'] = fam ~= 0 and fam or nil, stretchy = stretch}
+  return {[0] = 'mo', char, ['tex:family'] = fam ~= 0 and fam or nil, stretchy = stretch, [':node'] = acc}
 end
 
 local function kernel_to_table(kernel, cur_style)
@@ -164,7 +164,8 @@ local function kernel_to_table(kernel, cur_style)
     local result = {[0] = elem,
       char,
       ['tex:family'] = fam ~= 0 and fam or nil,
-      mathvariant = utf8.len(char) == 1 and elem == 'mi' and utf8.codepoint(char) < 0x10000 and 'normal' or nil
+      mathvariant = utf8.len(char) == 1 and elem == 'mi' and utf8.codepoint(char) < 0x10000 and 'normal' or nil,
+      [':node'] = kernel,
     }
     return result, result
   elseif id == sub_box_t then
@@ -172,7 +173,7 @@ local function kernel_to_table(kernel, cur_style)
       local result = to_text(kernel.list.head)
       return result, result
     else
-      local result = {[0] = 'mi', {[0] = 'mglyph', ['tex:box'] = kernel.list}}
+      local result = {[0] = 'mi', {[0] = 'mglyph', ['tex:box'] = kernel.list, [':node'] = kernel}}
       return result, result
     end
   elseif id == sub_mlist_t then
@@ -305,7 +306,7 @@ local function radical_to_table(radical, sub, cur_style)
   local elem
   if kind == 'radical' or kind == 'uradical' then
     -- FIXME: Check that this is really a square root
-    elem, core = {[0] = 'msqrt', nucleus}, nil
+    elem, core = {[0] = 'msqrt', nucleus, [':node'] = left[':node'], }, nil
   elseif kind == 'uroot' then
     -- FIXME: Check that this is really a root
     elem, core = {[0] = 'msqrt', nucleus, kernel_to_table(radical.degree)}, nil
@@ -331,6 +332,7 @@ local function fraction_to_table(fraction, sub, cur_style)
   local mfrac = {[0] = 'mfrac',
     linethickness = fraction.width and fraction.width == 0 and 0 or nil,
     bevelled = fraction.middle and "true" or nil,
+    [':node'] = fraction,
     num,
     denom,
   }

luamml-structelemwriter.lua

@@ -0,0 +1,83 @@
+local struct_begin = token.create'tag_struct_begin:n'
+local struct_end = token.create'tag_struct_end:'
+
+local mc_begin = token.create'tag_mc_begin:n'
+local mc_end = token.create'tag_mc_end:'
+
+local function escape_name(name)
+  return name
+end
+
+local function escape_string(str)
+  return str
+end
+
+local mathml_ns_obj
+local function get_mathml_ns_obj()
+  mathml_ns_obj = mathml_ns_obj or token.create'c__pdf_backend_object_tag/NS/mathml_int'.index
+  return mathml_ns_obj
+end
+
+local attribute_counter = 0
+local attributes = setmetatable({}, {__index = function(t, k)
+  attribute_counter = attribute_counter + 1
+  local attr_name = string.format('luamml_attr_%i', attribute_counter)
+  t[k] = attr_name
+  tex.runtoks(function()
+    -- tex.sprint(string.format('\\tagpdfsetup{newattribute={%s}{/O/NSO/NS %i 0 R',
+    --     attr_name, mathml_ns_obj or get_mathml_ns_obj()))
+    tex.sprint(string.format('\\tagpdfsetup{newattribute={%s}{/O/MathML-3',
+        attr_name))
+    tex.cprint(12, k)
+    tex.sprint'}}'
+  end)
+  return attr_name
+end})
+
+local mc_type = token.create'l__tag_mc_type_attr'.index
+local mc_cnt = token.create'l__tag_mc_cnt_attr'.index
+-- print('!!!', mc_type, mc_cnt)
+
+local attrs = {}
+local function write_elem(tree, indent)
+  if not tree[0] then print('ERR', require'inspect'(tree)) end
+  local i = 0
+  for attr, val in next, tree do if type(attr) == 'string' and not string.find(attr, ':') and attr ~= 'xmlns' then
+  -- for attr, val in next, tree do if type(attr) == 'string' and string.byte(attr) ~= 0x3A then
+    i = i + 1
+    attrs[i] = string.format('/%s(%s)', escape_name(attr), escape_string(val))
+  end end
+  table.sort(attrs)
+  local attr_name
+  if i == 0 then
+    tex.sprint{struct_begin, string.format('{tag=%s/mathml}', tree[0])}
+  else
+    tex.sprint{struct_begin, string.format('{tag=%s/mathml,attribute=%s}', tree[0], attributes[table.concat(attrs)])}
+  end
+  for j = 1, i do attrs[j] = nil end
+
+  if tree[':node'] then
+    local n = tree[':node']
+    tex.runtoks(function()
+      tex.sprint{mc_begin, string.format('{tag=%s}', tree[0])}
+      -- NOTE: This will also flush all previous sprint's... That's often annoying, but in this case actually intentional.
+    end)
+    node.set_attribute(tree[':node'], mc_type, tex.attribute[mc_type])
+    node.set_attribute(tree[':node'], mc_cnt, tex.attribute[mc_cnt])
+    tex.runtoks(function()
+      tex.sprint(mc_end)
+    end)
+  end
+  for _, elem in ipairs(tree) do
+    if type(elem) ~= 'string' then
+      write_elem(elem)
+    end
+  end
+  tex.runtoks(function()
+    tex.sprint(struct_end)
+  end)
+end
+
+return function(element)
+  return write_elem(element)
+end

luamml-tex.lua

@@ -5,6 +5,7 @@ local register_family = mlist_to_mml.register_family
 
 local mappings = require'luamml-legacy-mappings'
 local write_xml = require'luamml-xmlwriter'
+local write_struct = require'luamml-structelemwriter'
 
 local properties = node.get_properties_table()
 
@@ -67,6 +68,9 @@ luatexbase.add_to_callback('pre_mlist_to_hlist_filter', function(mlist, style)
   if flag & 2 == 0 then
     save_result(xml, style == 'display' or flag & 1 == 1)
   end
+  if flag & 8 == 8 then
+    write_struct(make_root(mlist_result, mlist_display and 0 or 2))
+  end
   if style == 'text' then
     local startmath = tex.nest.top.tail
     local props = properties[startmath]

luamml-xmlwriter.lua

@@ -18,7 +18,7 @@ local function write_elem(tree, indent)
   if not tree[0] then print('ERR', require'inspect'(tree)) end
   local escaped_name = escape_name(assert(tree[0]))
   local i = 0
-  for attr, val in next, tree do if type(attr) == 'string' then
+  for attr, val in next, tree do if type(attr) == 'string' and string.byte(attr) ~= 0x3A then
     i = i + 1
     attrs[i] = string.format(' %s="%s"', escape_name(attr), escape_text(val))
   end end

luamml.sty

@@ -17,6 +17,9 @@
 \cs_new:Nn \luamml_flag_alignment_right: {
   \int_set:Nn \l__luamml_flag_int { 6 }
 }
+\cs_new:Nn \luamml_flag_structelem: {
+  \int_set:Nn \l__luamml_flag_int { 8 }
+}
 
 \cs_new:Npn \__luamml_patch_package:nn #1 #2 {
   \@ifpackageloaded {#1} {#2} {

test_tex.tex

@@ -29,6 +29,7 @@
     2 & $else$
   \end{cases}
 \]
+\ShowMathMLObj
 \[
   x = \frac{-b \pm \sqrt{b^2-4ac}}{2a}.
 \]