Fix math mathvariants

packages/math/texlike.lua

@@ -42,8 +42,11 @@ local mathGrammar = function (_ENV)
          return ret
       end
    local group = P"{" * V"mathlist" * (P"}" + E("`}` expected"))
+   -- Simple amsmath-like \text command (no embedded math)
+   local textgroup = P"{" * C((1-P"}")^1) * (P"}" + E("`}` expected"))
    local element_no_infix =
       V"def" +
+      V"text" + -- Important: before command
       V"command" +
       group +
       V"argument" +
@@ -115,6 +118,11 @@ local mathGrammar = function (_ENV)
    sub = element_no_infix * _ * P"_" * _ * element_no_infix
    atom = natural + C(utf8code - S"\\{}%^_&") +
       (P"\\{" + P"\\}") / function (s) return string.sub(s, -1) end
+   text = (
+         P"\\text" *
+         Cg(parameters, "options") *
+         textgroup
+      )
    command = (
          P"\\" *
          Cg(ctrl_sequence_name, "command") *
@@ -252,6 +260,23 @@ local compileToStr = function (argEnv, mathlist)
    end
 end
 
+local function isBigOperator (tree)
+   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
+      return true
+   end
+   -- Case \mo{ops} where ops is registered as big operator (unicode-symbols)
+   -- E.g. \mo{∑) or \sum
+   if tree[1] and symbolDefaults[tree[1]] and symbolDefaults[tree[1]].atom == atomType.bigOperator then
+      return true
+   end
+   return false
+end
+
 local function compileToMathML_aux (_, arg_env, tree)
    if type(tree) == "string" then
       return tree
@@ -328,21 +353,13 @@ local function compileToMathML_aux (_, arg_env, tree)
       tree.options = {}
    -- Translate TeX-like sub/superscripts to `munderover` or `msubsup`,
    -- depending on whether the base is a big operator
-   elseif tree.id == "sup" and tree[1].command == "mo" and tree[1].atom == atomType.bigOperator then
+   elseif tree.id == "sup" and isBigOperator(tree[1]) then
       tree.command = "mover"
-   elseif tree.id == "sub" and tree[1].command == "mo" and symbolDefaults[tree[1][1]].atom == atomType.bigOperator then
+   elseif tree.id == "sub" and isBigOperator(tree[1]) then
       tree.command = "munder"
-   elseif
-      tree.id == "subsup"
-      and tree[1].command == "mo"
-      and symbolDefaults[tree[1][1]].atom == atomType.bigOperator
-   then
+   elseif tree.id == "subsup" and isBigOperator(tree[1])then
       tree.command = "munderover"
-   elseif
-      tree.id == "supsub"
-      and tree[1].command == "mo"
-      and symbolDefaults[tree[1][1]].atom == atomType.bigOperator
-   then
+   elseif tree.id == "supsub" and isBigOperator(tree[1]) then
       tree.command = "munderover"
       local tmp = tree[2]
       tree[2] = tree[3]
@@ -365,6 +382,8 @@ local function compileToMathML_aux (_, arg_env, tree)
          return compileToMathML_aux(nil, compiledArgs, tree[1])
       end)
       return nil
+   elseif tree.id == "text" then
+      tree.command = "mtext"
    elseif tree.id == "command" and commands[tree.command] then
       local argTypes = commands[tree.command][1]
       local cmdFun = commands[tree.command][2]
@@ -425,7 +444,7 @@ local function printMathML (tree)
    if tree.options then
       local options = {}
       for k, v in pairs(tree.options) do
-         table.insert(options, k .. "=" .. v)
+         table.insert(options, k .. "=" .. tostring(v))
       end
       if #options > 0 then
          result = result .. "[" .. table.concat(options, ", ") .. "]"
@@ -480,6 +499,21 @@ compileToMathML(
   \def{bi}{\mi[mathvariant=bold-italic]{#1}}
   \def{dsi}{\mi[mathvariant=double-struck]{#1}}
 
+  \def{lim}{\mo[atom=big]{lim}}
+
+  % From amsmath:
+  \def{to}{\mo[atom=bin]{→}}
+  \def{gcd}{\mo[atom=big]{gcd}}
+  \def{sup}{\mo[atom=big]{sup}}
+  \def{inf}{\mo[atom=big]{inf}}
+  \def{max}{\mo[atom=big]{max}}
+  \def{min}{\mo[atom=big]{min}}
+  % Those use U+202F NARROW NO-BREAK SPACE in their names
+  \def{limsup}{\mo[atom=big]{lim sup}}
+  \def{liminf}{\mo[atom=big]{lim inf}}
+  \def{projlim}{\mo[atom=big]{proj lim}}
+  \def{injlim}{\mo[atom=big]{inj lim}}
+
   % Standard spaces gleaned from plain TeX
   \def{thinspace}{\mspace[width=thin]}
   \def{negthinspace}{\mspace[width=-thin]}
@@ -496,9 +530,44 @@ compileToMathML(
   \def{quad}{\mspace[width=1em]}
   \def{qquad}{\mspace[width=2em]}
 
+  % MathML says a single-character identifier must be in italic by default.
+  % TeX however has the following Greek capital macros rendered in upright shape.
+  % It so common that you've probably never seen Γ(x) written with an italic gamma.
+  \def{Gamma}{\mi[mathvariant=normal]{Γ}}
+  \def{Delta}{\mi[mathvariant=normal]{Δ}}
+  \def{Theta}{\mi[mathvariant=normal]{Θ}}
+  \def{Lambda}{\mi[mathvariant=normal]{Λ}}
+  \def{Xi}{\mi[mathvariant=normal]{Ξ}}
+  \def{Pi}{\mi[mathvariant=normal]{Π}}
+  \def{Sigma}{\mi[mathvariant=normal]{Σ}}
+  \def{Upsilon}{\mi[mathvariant=normal]{Υ}}
+  \def{Phi}{\mi[mathvariant=normal]{Φ}}
+  \def{Psi}{\mi[mathvariant=normal]{Ψ}}
+  \def{Omega}{\mi[mathvariant=normal]{Ω}}
+  % Some calligraphic (script), fraktur, double-struck styles:
+  % Convenience for compatibility with LaTeX.
+  \def{mathcal}{\mi[mathvariant=script]{#1}}
+  \def{mathfrak}{\mi[mathvariant=fraktur]{#1}}
+  \def{mathbb}{\mi[mathvariant=double-struck]{#1}}
+  % Some style-switching commands for compatibility with LaTeX math.
+  % Caveat emptor: LaTeX would allow these to apply to a whole formula.
+  % We can't do that in MathML, as mathvariant applies to token elements only.
+  % Also note that LaTeX and related packages may have many more such commands.
+  % We only provide a few common ('historical') ones here.
+  \def{mathrm}{\mi[mathvariant=normal]{#1}}
+  \def{mathbf}{\mi[mathvariant=bold]{#1}}
+  \def{mathit}{\mi[mathvariant=italic]{#1}}
+  \def{mathsf}{\mi[mathvariant=sans-serif]{#1}}
+  \def{mathtt}{\mi[mathvariant=monospace]{#1}}
+
   % Modulus operator forms
   \def{bmod}{\mo{mod}}
   \def{pmod}{\quad(\mo{mod} #1)}
+
+  % Phantom commands from TeX/LaTeX
+  \def{phantom}{\mphantom{#1}}
+  \def{hphantom}{\mphantom[special=h]{#1}}
+  \def{vphantom}{\mphantom[special=v]{#1}}
 ]==],
    })
 )

packages/math/typesetter.lua

@@ -1,6 +1,8 @@
 -- Interpret a MathML or TeX-like AST, typeset it and add it to the output.
 local b = require("packages.math.base-elements")
 local syms = require("packages.math.unicode-symbols")
+local mathvariants = require("packages.math.unicode-mathvariants")
+local mathVariantToScriptType, scriptType = mathvariants.mathVariantToScriptType, mathvariants.scriptType
 
 -- Shorthands for atom types, used in the `atom` command option
 local atomTypeShort = {
@@ -41,17 +43,22 @@ function ConvertMathML (_, content)
       return b.stackbox("V", convertChildren(content))
    elseif content.command == "mrow" then
       return b.stackbox("H", convertChildren(content))
+   elseif content.command == "mphantom" then
+      -- MathML's standard mphantom corresponds to TeX's \phantom only.
+      -- Let's support a special attribute "h" or "v" for TeX-like \hphantom or \vphantom.
+      local special = content.options.special
+      return b.phantom(convertChildren(content), special)
    elseif content.command == "mi" then
-      local script = content.options.mathvariant and b.mathVariantToScriptType(content.options.mathvariant)
+      local script = content.options.mathvariant and mathVariantToScriptType(content.options.mathvariant)
       local text = content[1]
       if type(text) ~= "string" then
-         SU.error("mi command contains " .. text .. ", which is not text")
+         SU.error("mi command contains content which is not text")
       end
-      script = script or (luautf8.len(text) == 1 and b.scriptType.italic or b.scriptType.upright)
+      script = script or (luautf8.len(text) == 1 and scriptType.italic or scriptType.upright)
       return b.text("identifier", {}, script, text)
    elseif content.command == "mo" then
-      local script = content.options.mathvariant and b.mathVariantToScriptType(content.options.mathvariant)
-         or b.scriptType.upright
+      local script = content.options.mathvariant and mathVariantToScriptType(content.options.mathvariant)
+         or scriptType.upright
       local text = content[1]
       local attributes = {}
       if syms.symbolDefaults[text] then
@@ -67,15 +74,15 @@ function ConvertMathML (_, content)
          end
       end
       if type(text) ~= "string" then
-         SU.error("mo command contains " .. text .. ", which is not text")
+         SU.error("mo command contains content which is not text")
       end
       return b.text("operator", attributes, script, text)
    elseif content.command == "mn" then
-      local script = content.options.mathvariant and b.mathVariantToScriptType(content.options.mathvariant)
-         or b.scriptType.upright
+      local script = content.options.mathvariant and mathVariantToScriptType(content.options.mathvariant)
+         or scriptType.upright
       local text = content[1]
       if type(text) ~= "string" then
-         SU.error("mn command contains " .. text .. ", which is not text")
+         SU.error("mn command contains content which is not text")
       end
       if string.sub(text, 1, 1) == "-" then
          text = "−" .. string.sub(text, 2)
@@ -129,13 +136,29 @@ function ConvertMathML (_, content)
       local children = convertChildren(content)
       -- "The <msqrt> element generates an anonymous <mrow> box called the msqrt base
       return b.sqrt(b.stackbox("H", children))
+   elseif content.command == "mroot" then
+      local children = convertChildren(content)
+      return b.sqrt(children[1], children[2])
    elseif content.command == "mtable" or content.command == "table" then
       local children = convertChildren(content)
       return b.table(children, content.options)
    elseif content.command == "mtr" then
       return b.mtr(convertChildren(content))
    elseif content.command == "mtd" then
       return b.stackbox("H", convertChildren(content))
+   elseif content.command == "mtext" or content.command == "ms" then
+      if #content > 1 then
+         SU.error("Wrong number of children in " .. content.command .. ": " .. #content)
+      end
+      local text = content[1] or "" -- empty mtext is allowed, and found in examples...
+      if type(text) ~= "string" then
+         SU.error(content.command .. " command contains content which is not text")
+      end
+      -- MathML Core 3.2.1.1 Layout of <mtext> has some wording about forced line breaks
+      -- and soft wrap opportunities: ignored here.
+      -- 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+", " "))
    else
       SU.error("Unknown math command " .. content.command)
    end