require "nokogiri"
require "htmlentities"
require "pp"

module MathML2AsciiMath
  def self.m2a(xml)
    normalized = xml

    # &:noblanks skips non-significant whitespaces in MathML
    docxml = Nokogiri::XML.parse(normalized, &:noblanks)

    # Get rid of things like
    #           <mtext>&#x2009;</mtext>
    parse(docxml.root).gsub(/[[:blank:]]/, " ").unicode_normalize.squeeze(" ")
  end

  def self.encodechars(xml)
    xml.gsub(/\u03b1/, "alpha")
      .gsub(/\u03b2/, "beta")
      .gsub(/\u03b3/, "gamma")
      .gsub(/\u0393/, "Gamma")
      .gsub(/\u03b4/, "delta")
      .gsub(/\u0394/, "Delta")
      .gsub(/\u2206/, "Delta")
      .gsub(/\u03b5/, "epsilon")
      .gsub(/\u025b/, "varepsilon")
      .gsub(/\u03b6/, "zeta")
      .gsub(/\u03b7/, "eta")
      .gsub(/\u03b8/, "theta")
      .gsub(/\u0398/, "Theta")
      .gsub(/\u03d1/, "vartheta")
      .gsub(/\u03b9/, "iota")
      .gsub(/\u03ba/, "kappa")
      .gsub(/\u03bb/, "lambda")
      .gsub(/\u039b/, "Lambda")
      .gsub(/\u03bc/, "mu")
      .gsub(/\u03bd/, "nu")
      .gsub(/\u03be/, "xi")
      .gsub(/\u039e/, "Xi")
      .gsub(/\u03c0/, "pi")
      .gsub(/\u03a0/, "Pi")
      .gsub(/\u03c1/, "rho")
      .gsub(/\u03c2/, "beta")
      .gsub(/\u03c3/, "sigma")
      .gsub(/\u03a3/, "Sigma")
      .gsub(/\u03c4/, "tau")
      .gsub(/\u03c5/, "upsilon")
      .gsub(/\u03c6/, "phi")
      .gsub(/\u03a6/, "Phi")
      .gsub(/\u03d5/, "varphi")
      .gsub(/\u03c7/, "chi")
      .gsub(/\u03c8/, "psi")
      .gsub(/\u03a8/, "Psi")
      .gsub(/\u03c9/, "omega")
      .gsub(/\u03a9/, "omega")
      .gsub(/\u22c5/, "*")
      .gsub(/\u2219/, "*")
      .gsub(/\u00b7/, "*")
      .gsub(/\u2217/, "**")
      .gsub(/\u22c6/, "***")
      .gsub(/\//, "//")
      .gsub(/\\/, "\\\\")
      .gsub(/\u00d7/, "xx")
      .gsub(/\u22c9/, "|><")
      .gsub(/\u22ca/, "><|")
      .gsub(/\u22c8/, "|><|")
      .gsub(/\u00f7/, "-:")
      .gsub(/\u2218/, "@")
      .gsub(/\u2295/, "o+")
      .gsub(/\u2a01/, "o+")
      .gsub(/\u2297/, "ox")
      .gsub(/\u2299/, "o.")
      .gsub(/\u2211/, "sum")
      .gsub(/\u220f/, "prod")
      .gsub(/\u2227/, "^^")
      .gsub(/\u22c0/, "^^^")
      .gsub(/\u2228/, "vv")
      .gsub(/\u22c1/, "vvv")
      .gsub(/\u2229/, "nn")
      .gsub(/\u22c2/, "nnn")
      .gsub(/\u222a/, "uu")
      .gsub(/\u22c3/, "uuu")
      .gsub(/\u2260/, "!=")
      .gsub(/\u2264/, "<=")
      .gsub(/\u2265/, ">=")
      .gsub(/\u227a/, "-<")
      .gsub(/\u227b/, ">-")
      .gsub(/\u2aaf/, "-<=")
      .gsub(/\u2ab0/, ">-=")
      .gsub(/\u2208/, "in")
      .gsub(/\u2209/, "!in")
      .gsub(/\u2282/, "sub")
      .gsub(/\u2283/, "sup")
      .gsub(/\u2286/, "sube")
      .gsub(/\u2287/, "supe")
      .gsub(/\u2261/, "-=")
      .gsub(/\u2245/, "~=")
      .gsub(/\u2248/, "~~")
      .gsub(/\u221d/, "prop")
      .gsub(/\u00ac/, "not")
      .gsub(/\u21d2/, "=>")
      .gsub(/\u21d4/, "<=>")
      .gsub(/\u2200/, "AA")
      .gsub(/\u2203/, "EE")
      .gsub(/\u22a5/, "_|_")
      .gsub(/\u22a4/, "TT")
      .gsub(/\u22a2/, "|--")
      .gsub(/\u22a8/, "|==")
      .gsub(/\u22a8/, "|==")
      .gsub(/\u2329/, "(:")
      .gsub(/\u232a/, ":)")
      .gsub(/\u2329/, "<<")
      .gsub(/\u27e8/, "<<")
      .gsub(/\u232a/, ">>")
      .gsub(/\u27e9/, ">>")
      .gsub(/\u222b/, "int")
      .gsub(/\u222e/, "oint")
      .gsub(/\u2202/, "del")
      .gsub(/\u2207/, "grad")
      .gsub(/\u00b1/, "+-")
      .gsub(/\u2205/, "O/")
      .gsub(/\u221e/, "oo")
      .gsub(/\u2135/, "aleph")
      .gsub(/\u2234/, ":.")
      .gsub(/\u2235/, ":'")
      .gsub(/\u2220/, "/_")
      .gsub(/\u25b3/, "/_\\")
      .gsub(/\u2032/, "'")
      .gsub(/~/, "tilde")
      .gsub(/\u00a0\u00a0\u00a0\u00a0/, "qquad")
      .gsub(/\u00a0\u00a0/, "quad")
      .gsub(/\u00a0/, "\\ ")
      .gsub(/\u2322/, "frown")
      .gsub(/\u00a0/, "quad")
      .gsub(/\u22ef/, "cdots")
      .gsub(/\u22ee/, "vdots")
      .gsub(/\u22f1/, "ddots")
      .gsub(/\u22c4/, "diamond")
      .gsub(/\u25a1/, "square")
      .gsub(/\u230a/, "|__")
      .gsub(/\u230b/, "__|")
      .gsub(/\u2308/, "|~")
      .gsub(/\u2309/, "~|")
      .gsub(/\u2102/, "CC")
      .gsub(/\u2115/, "NN")
      .gsub(/\u211a/, "QQ")
      .gsub(/\u211d/, "RR")
      .gsub(/\u2124/, "ZZ")
      .gsub(/\u2191/, "uarr")
      .gsub(/\u2193/, "darr")
      .gsub(/\u2190/, "larr")
      .gsub(/\u2194/, "harr")
      .gsub(/\u21d2/, "rArr")
      .gsub(/\u21d0/, "lArr")
      .gsub(/\u21d4/, "hArr")
      .gsub(/\u2192/, "->")
      .gsub(/\u21a3/, ">->")
      .gsub(/\u21a0/, "->>")
      .gsub(/\u2916/, ">->>")
      .gsub(/\u21a6/, "|->")
      .gsub(/\u2026/, "...")
      .gsub(/\u2212/, "-")
      .gsub(/\u2061/, "") # function application
      .gsub(/\u2751/, "square")
      .gsub(/[\u2028\u2029]/, " ") # normalize thin spaces like \u2009, \u2008
  end

  def self.join_parsed_children(children, delimiter = " ")
    children.map do |n|
      parse(n).strip
    end.join(delimiter)
  end

  def self.parse(node)
    out = ""
    if node.text?
      return encodechars(HTMLEntities.new.decode(node.text))
    end

    case node.name.sub(/^[^:]*:/, "")
    when "math"
      join_parsed_children(node.elements)

    when "annotation"
      ""

    when "semantics"
      join_parsed_children(node.elements)

    when "mrow"
      out = join_parsed_children(node.elements)
      if %w[mfrac msub munder munderover]
        .include? node.parent.name.sub(/^[^:]*:/, "")
        out = "(#{out})"
      end
      out

    when "mfenced"
      sym_open = node["open"] || "("
      sym_close = node["close"] || ")"

      separator = "," # TODO currently ignore the supplied separators
      out = join_parsed_children(node.elements, separator)
      "#{sym_open}#{out}#{sym_close}"

    when "msqrt"
      "sqrt(#{join_parsed_children(node.elements)})"

    when "mfrac"
      "(#{parse(node.elements[0])})/(#{parse(node.elements[1])})"

    when "msup"
      sup = parse(node.elements[1])
      sup = "(#{sup})" unless sup.length == 1
      op = parse(node.elements[0]).gsub(/ $/, "")
      "#{op}^#{sup}"

    when "msub"
      sub = parse(node.elements[1])
      sub = "(#{sub})" unless sub.length == 1
      op = parse(node.elements[0]).gsub(/ $/, "")
      "#{op}_#{sub}"

    when "munderover", "msubsup"
      sub = parse(node.elements[1])
      sub = "(#{sub})" unless sub.length == 1
      sup = parse(node.elements[2])
      sup = "(#{sup})" unless sup.length == 1
      op = parse(node.elements[0]).gsub(/ $/, "")
      "#{op}_#{sub}^#{sup}"

    when "munder"
      elem1 = parse(node.elements[1]).strip
      accent = case elem1
               when "\u0332" then "ul"
               when "\u23df" then "ubrace"
               else
                 "underset"
               end

      if accent == "underset"
        "underset(#{elem1})(#{parse(node.elements[0])})"
      else
        "#{accent} #{parse(node.elements[0])}"
      end

    when "mover"
      elem1 = parse(node.elements[1]).strip
      accent = case elem1
               when "\u005e" then "hat"
               when "\u00af" then "bar"
               # when "\u2192" then "vec"
               when "->" then "vec"
               when "." then "dot"
               when ".." then "ddot"
               when "\u23de" then "obrace"
               else
                 "overset"
               end

      if accent == "overset"
        "overset(#{elem1})(#{parse(node.elements[0])})"
      else
        "#{accent} #{parse(node.elements[0])}"
      end

    when "mtable"
      "[#{join_parsed_children(node.elements, ',')}]"

    when "mtr"
      "[#{join_parsed_children(node.elements, ',')}]"

    when "mtd"
      join_parsed_children(node.elements, ",")

    when "mn", "mtext"
      join_parsed_children(node.children, "")

    when "mi"
      # FIXME: What does this comment have to do with Word?
      # mi is not meant to have space around it,
      # but Word is conflating operators and operands
      join_parsed_children(node.children)

      # FIXME: Why do we need to add extra spaces?
      # out = " #{out} " if /[^a-zA-Z0-9',]|[a-z][a-z]/.match out

    when "mo"
      out = join_parsed_children(node.children)
      out = " #{out} " unless node["fence"]
      out

    when "mstyle"
      join_parsed_children(node.children)

    else
      "<math xmlns=\"http://www.w3.org/1998/Math/MathML\">" +
        node.to_xml +
        "</math>"

    end
  end
end