Merge pull request #2151 from Omikhleia/math-more-fixes-and-features

Math more fixes and features

packages/math/texlike.lua

@@ -260,23 +260,40 @@ local compileToStr = function (argEnv, mathlist)
    end
 end
 
-local function isBigOperator (tree)
+local function isOperatorKind (tree, typeOfAtom, typeOfSymbol)
+   if not tree then
+      return false -- safeguard
+   end
    if tree.command ~= "mo" then
       return false
    end
    -- Case \mo[atom=big]{ops}
    -- E.g. \mo[atom=big]{lim}
-   if tree.options and tree.options.atom == "big" then
+   if tree.options and tree.options.atom == typeOfAtom then
       return true
    end
-   -- Case \mo{ops} where ops is registered as big operator (unicode-symbols)
+   -- Case \mo{ops} where ops is registered with the resquested type
    -- E.g. \mo{∑) or \sum
-   if tree[1] and symbolDefaults[tree[1]] and symbolDefaults[tree[1]].atom == atomType.bigOperator then
+   if tree[1] and symbolDefaults[tree[1]] and symbolDefaults[tree[1]].atom == typeOfSymbol then
       return true
    end
    return false
 end
 
+local function isBigOperator (tree)
+   return isOperatorKind(tree, "big", atomType.bigOperator)
+end
+local function isCloseOperator (tree)
+   return isOperatorKind(tree, "close", atomType.closeSymbol)
+end
+local function isOpeningOperator (tree)
+   return isOperatorKind(tree, "open", atomType.openingSymbol)
+end
+
+local function isAccentSymbol (symbol)
+   return symbolDefaults[symbol] and symbolDefaults[symbol].atom == atomType.accentSymbol
+end
+
 local function compileToMathML_aux (_, arg_env, tree)
    if type(tree) == "string" then
       return tree
@@ -323,12 +340,56 @@ local function compileToMathML_aux (_, arg_env, tree)
       -- Turn mathlist into `mrow` except if it has exactly one `mtr` or `mtd`
       -- child.
       -- Note that `def`s have already been compiled away at this point.
-      if #tree == 1 and (tree[1].command == "mtr" or tree[1].command == "mtd") then
-         return tree[1]
+      if #tree == 1 then
+         if tree[1].command == "mtr" or tree[1].command == "mtd" then
+            return tree[1]
+         else
+            tree.command = "mrow"
+         end
       else
+         -- Re-wrap content from opening to closing operator in an implicit mrow,
+         -- so stretchy operators apply to the correct span of content.
+         local children = {}
+         local stack = {}
+         for _, child in ipairs(tree) do
+            if isOpeningOperator(child) then
+               table.insert(stack, children)
+               local mrow = {
+                  command = "mrow",
+                  options = {},
+                  child,
+               }
+               table.insert(children, mrow)
+               children = mrow
+            elseif isCloseOperator(child) then
+               table.insert(children, child)
+               if #stack > 0 then
+                  children = table.remove(stack)
+               end
+            elseif
+               (child.command == "msubsup" or child.command == "msub" or child.command == "msup")
+               and isCloseOperator(child[1]) -- child[1] is the base
+            then
+               if #stack > 0 then
+                  -- Special case for closing operator with sub/superscript:
+                  -- (....)^i must be interpreted as {(....)}^i, not as (...{)}^i
+                  -- Push the closing operator into the mrow
+                  table.insert(children, child[1])
+                  -- Move the mrow into the msubsup, replacing the closing operator
+                  child[1] = children
+                  -- And insert the msubsup into the parent
+                  children = table.remove(stack)
+                  children[#children] = child
+               else
+                  table.insert(children, child)
+               end
+            else
+               table.insert(children, child)
+            end
+         end
+         tree = #stack > 0 and stack[1] or children
          tree.command = "mrow"
       end
-      tree.command = "mrow"
    elseif tree.id == "atom" then
       local codepoints = {}
       for _, cp in luautf8.codes(tree[1]) do
@@ -424,7 +485,36 @@ local function compileToMathML_aux (_, arg_env, tree)
       return res
    elseif tree.id == "command" and symbols[tree.command] then
       local atom = { id = "atom", [1] = symbols[tree.command] }
-      tree = compileToMathML_aux(nil, arg_env, atom)
+      if isAccentSymbol(symbols[tree.command]) and #tree > 0 then
+         -- LaTeX-style accents \vec{v} = <mover accent="true"><mi>v</mi><mo>→</mo></mover>
+         local accent = {
+            id = "command",
+            command = "mover",
+            options = {
+               accent = "true",
+            },
+         }
+         accent[1] = compileToMathML_aux(nil, arg_env, tree[1])
+         accent[2] = compileToMathML_aux(nil, arg_env, atom)
+         tree = accent
+      elseif #tree > 0 then
+         -- Play cool with LaTeX-style commands that don't take arguments:
+         -- Edge case for non-accent symbols so we don't loose bracketed groups
+         -- that might have been seen as command arguments.
+         -- Ex. \langle{x}\rangle (without space after \langle)
+         local sym = compileToMathML_aux(nil, arg_env, atom)
+         -- Compile all children in-place
+         for i, child in ipairs(tree) do
+            tree[i] = compileToMathML_aux(nil, arg_env, child)
+         end
+         -- Insert symbol at the beginning,
+         -- And add a wrapper mrow to be unwrapped in the parent.
+         table.insert(tree, 1, sym)
+         tree.command = "mrow"
+         tree.id = "wrapper"
+      else
+         tree = compileToMathML_aux(nil, arg_env, atom)
+      end
    elseif tree.id == "argument" then
       if arg_env[tree.index] then
          return arg_env[tree.index]
@@ -566,8 +656,9 @@ compileToMathML(
 
   % Phantom commands from TeX/LaTeX
   \def{phantom}{\mphantom{#1}}
-  \def{hphantom}{\mphantom[special=h]{#1}}
-  \def{vphantom}{\mphantom[special=v]{#1}}
+  \def{hphantom}{\mpadded[height=0, depth=0]{\mphantom{#1}}}
+  \def{vphantom}{\mpadded[width=0]{\mphantom{#1}}}
+  %\mphantom[special=v]{#1}}}
 ]==],
    })
 )

packages/math/typesetter.lua

@@ -34,13 +34,24 @@ local function convertChildren (tree)
    return mboxes
 end
 
+local function convertFirstChild (tree)
+   -- We need to loop until the first non-nil box is found, because
+   -- we may have blank lines in the tree.
+   for _, n in ipairs(tree) do
+      local box = ConvertMathML(nil, n)
+      if box then
+         return box
+      end
+   end
+end
+
 -- convert MathML into mbox
 function ConvertMathML (_, content)
    if content == nil or content.command == nil then
       return nil
    end
    if content.command == "math" or content.command == "mathml" then -- toplevel
-      return b.stackbox("V", convertChildren(content))
+      return b.stackbox("H", convertChildren(content))
    elseif content.command == "mrow" then
       return b.stackbox("H", convertChildren(content))
    elseif content.command == "mphantom" then
@@ -61,11 +72,16 @@ function ConvertMathML (_, content)
          or scriptType.upright
       local text = content[1]
       local attributes = {}
+      -- Attributes from the (default) oerator table
       if syms.symbolDefaults[text] then
          for attribute, value in pairs(syms.symbolDefaults[text]) do
             attributes[attribute] = value
          end
       end
+      -- Overwrite with attributes from the element
+      for attribute, value in pairs(content.options) do
+         attributes[attribute] = value
+      end
       if content.options.atom then
          if not atomTypeShort[content.options.atom] then
             SU.error("Unknown atom type " .. content.options.atom)
@@ -131,7 +147,7 @@ function ConvertMathML (_, content)
       if #children ~= 2 then
          SU.error("Wrong number of children in mfrac: " .. #children)
       end
-      return b.fraction(children[1], children[2])
+      return b.fraction(content.options, children[1], children[2])
    elseif content.command == "msqrt" then
       local children = convertChildren(content)
       -- "The <msqrt> element generates an anonymous <mrow> box called the msqrt base
@@ -159,6 +175,17 @@ function ConvertMathML (_, content)
       -- There's also some explanations about CSS, italic correction etc. which we ignore too.
       text = text:gsub("[\n\r]", " ")
       return b.text("string", {}, scriptType.upright, text:gsub("%s+", " "))
+   elseif content.command == "maction" then
+      -- MathML Core 3.6: display as mrow, ignoring all but the first child
+      return b.stackbox("H", { convertFirstChild(content) })
+   elseif content.command == "mstyle" then
+      -- It's an mrow, but with some style attributes that we ignore.
+      SU.warn("MathML mstyle is not fully supported yet")
+      return b.stackbox("H", convertChildren(content))
+   elseif content.command == "mpadded" then
+      -- MathML Core 3.3.6.1: The <mpadded> element generates an anonymous <mrow> box
+      -- called the "impadded inner box"
+      return b.padded(content.options, b.stackbox("H", convertChildren(content)))
    else
       SU.error("Unknown math command " .. content.command)
    end

tests/bug-1495-inline-math-layout.expected

@@ -65,13 +65,11 @@ T	30 w=6.4800	(nil)
 Mx 	217.4262
 T	68 80 84 w=13.2200	(nil)
 Mx 	230.6462
-Mx 	2.8200
-T	9 a=3.5600	(nil)
+T	9 w=3.5600	(nil)
 Mx 	234.2062
 T	3421 w=5.9100	(nil)
 Mx 	240.1162
-Mx 	2.8200
-T	10 a=3.5600	(nil)
+T	10 w=3.5600	(nil)
 Mx 	247.3974
 T	30 w=6.4800	(nil)
 Mx 	257.5986
@@ -88,53 +86,49 @@ My 	52.6701
 T	87 68 72 w=12.6514	(tae)
 Mx 	27.5333
 T	17 w=2.2900	(.)
-Mx 	32.4913
+Mx 	32.7068
 T	48 68 72 70 72 81 68 86 w=40.2295	(Maecenas)
-Mx 	75.3888
+Mx 	75.8198
 T	89 72 81 72 81 68 87 76 86 w=39.8096	(venenatis)
-Mx 	117.8663
+Mx 	118.5129
 T	88 79 87 85 76 70 72 86 w=30.9717	(ultrices)
-Mx 	148.8380
+Mx 	149.4846
 T	17 w=2.2900	(.)
-Mx 	153.7961
+Mx 	154.6581
 T	51 75 68 86 72 79 79 88 86 w=38.3984	(Phasellus)
-Mx 	194.8625
+Mx 	195.9401
 Set font 	Libertinus Math;10;400;Regular;normal;;;LTR
 T	2750 w=3.8100	(nil)
-Mx 	198.6725
+Mx 	199.7501
 My 	49.5701
 Set font 	Libertinus Math;8;400;Regular;normal;;;LTR
 T	2753 w=1.8320	(nil)
-Mx 	200.5045
+Mx 	201.5821
 T	3413 w=3.5152	(nil)
-Mx 	207.2916
+Mx 	208.9129
 My 	52.6701
 Set font 	Libertinus Math;10;400;Regular;normal;;;LTR
 T	30 w=6.4800	(nil)
-Mx 	216.6135
+Mx 	218.7785
 T	68 80 84 w=13.2200	(nil)
-Mx 	229.8335
-Mx 	3.0000
-T	3798 a=4.2100	(nil)
-Mx 	234.0435
+Mx 	231.9985
+T	9 w=3.5600	(nil)
+Mx 	235.5585
 T	3413 w=4.5600	(nil)
-Mx 	238.6035
-Mx 	2.9900
-T	3799 a=4.2100	(nil)
-Mx 	245.0614
+Mx 	240.1185
+T	10 w=3.5600	(nil)
+Mx 	246.1439
 T	12 w=6.4800	(nil)
-Mx 	253.7892
+Mx 	255.0892
 T	2753 w=2.2900	(nil)
-Mx 	257.7459
+Mx 	259.0459
 T	84 74 79 w=12.0300	(nil)
-Mx 	269.7759
-Mx 	3.0000
-T	3798 a=4.2100	(nil)
-Mx 	273.9859
+Mx 	271.0759
+T	9 w=3.5600	(nil)
+Mx 	274.6359
 T	3413 w=4.5600	(nil)
-Mx 	278.5459
-Mx 	2.9900
-T	3799 a=4.2100	(nil)
+Mx 	279.1959
+T	10 w=3.5600	(nil)
 Mx 	14.8819
 My 	64.6701
 Set font 	Gentium Plus;10;400;;normal;;;LTR