lua-htmlparser/ElementNode.lua

require("luarocks.loader")
local Set = require("Set")

local ElementNode = {}
ElementNode.mt = {__index = ElementNode}
function ElementNode:new(nameortext, node, descend, openstart, openend)
  local instance = {
    name = nameortext,
    level = 0,
    parent = nil,
    root = nil,
    nodes = {},
    _openstart = openstart, _openend = openend,
    _closestart = openstart, _closeend = openend,
    attributes = {},
    id = nil,
    classes = {},
    deepernodes = Set:new(),
    deeperelements = {}, deeperattributes = {}, deeperids = {}, deeperclasses = {}
  }
  if not node then
    instance.name = "root"
    instance.root = instance
    instance._text = nameortext
    local length = string.len(nameortext)
    instance._openstart, instance._openend = 1, length
    instance._closestart, instance._closeend = 1, length
  elseif descend then
    instance.root = node.root
    instance.parent = node
    instance.level = node.level + 1
    table.insert(node.nodes, instance)
  else
    instance.root = node.root
    instance.parent = node.parent
    instance.level = node.level
    table.insert(node.parent.nodes, instance)
  end
  return setmetatable(instance, ElementNode.mt)
end

function ElementNode:gettext()
  return string.sub(self.root._text, self._openstart, self._closeend)
end

function ElementNode:getcontent()
  return string.sub(self.root._text, self._openend + 1, self._closestart - 1)
end

function ElementNode:addattribute(k, v)
  self.attributes[k] = v
  if string.lower(k) == "id" then
    self.id = v
  end
  -- class attribute contains "space-separated tokens", each of which we'd like quick access to
  if string.lower(k) == "class" then
    for class in string.gmatch(v, "%S+") do
      table.insert(self.classes, class)
    end
  end
end

local function insert(table, name, node)
  table[name] = table[name] or Set:new()
  table[name]:add(node)
end

function ElementNode:close(closestart, closeend)
  if closestart and closeend then
    self._closestart, self._closeend = closestart, closeend
  end
  -- inform hihger level nodes about this element's existence in their branches
  local node = self
  while true do
    node = node.parent
    if not node then break end
    node.deepernodes:add(self)
    insert(node.deeperelements, self.name, self)
    for k in pairs(self.attributes) do
      insert(node.deeperattributes, k, self)
    end
    if self.id then
      insert(node.deeperids, self.id, self)
    end
    for _,v in ipairs(self.classes) do
      insert(node.deeperclasses, v, self)
    end
  end
end

local function escape(s)
  -- escape all ^, $, (, ), %, ., [, ], *, +, - , and ? with a % prefix
  return string.gsub(s, "([%^%$%(%)%%%.%[%]%*%+%-%?])", "%%" .. "%1")
end

local function select(self, s)
  if not s or type(s) ~= "string" or s == "" then return Set:new() end
  local sets = {[""]  = self.deeperelements, ["["] = self.deeperattributes,
                ["#"] = self.deeperids,      ["."] = self.deeperclasses}
  local function match(t, w)
    local m, v
    if t == "[" then w, m, v = string.match(w, 
        "([^=|%*~%$!%^]+)" .. -- w = 1 or more characters up to a possible "=", "|", "*", "~", "$", "!", or "^"
        "([|%*~%$!%^]?)" ..   -- m = an optional "|", "*", "~", "$", "!", or "^", preceding the optional "="
        "=?" ..               -- an optional uncaptured "="
        "(.*)"                -- v = anything following the "=", or else ""
      )
    end
    local matched = Set:new(sets[t][w])
    -- attribute value selectors
    if v and v ~= "" then
      v = string.sub(v, 2, #v - 1) -- strip quotes
      for node in pairs(matched) do
        local a = node.attributes[w]
        -- equals
        if m == "" and a ~= v then matched:remove(node)
        -- not equals
        elseif m == "!" and a == v then matched:remove(node)
        -- prefix
        elseif m =="|" and string.match(a, "^[^-]*") ~= v then matched:remove(node)
        -- contains
        elseif m =="*" and string.match(a, escape(v)) ~= v then matched:remove(node)
        -- word
        elseif m =="~" then matched:remove(node)
          for word in string.gmatch(a, "%S+") do
            if word == v then matched:add(node) break end
          end
        -- starts with
        elseif m =="^" and string.match(a, "^" .. escape(v)) ~= v then matched:remove(node)
        -- ends with
        elseif m =="$" and string.match(a, escape(v) .. "$") ~= v then matched:remove(node)
        end
      end -- for node
    end -- if v
    return matched
  end

  local subjects, resultset, childrenonly = Set:new({self})
  for part in string.gmatch(s, "%S+") do
    if part == ">" then childrenonly = true goto nextpart end
    resultset = Set:new()
    for subject in pairs(subjects) do
      local star = subject.deepernodes
      if childrenonly then star = Set:new(subject.nodes) childrenonly = false end
      resultset = resultset + star
    end
    if part == "*" then goto nextpart end
    local excludes, filter = Set:new()
    for t, w in string.gmatch(part,
      "([:%[#.]?)" ..        -- t = an optional :, [, #, or .
      "([^:%(%[#.%]%)]+)" .. -- w = 1 or more of anything not :, (, [, #, ., ], or )
      "%]?%)?"               -- followed by an uncaptured optional ] and/or )
    ) do
      if t == ":" then filter = w goto nextw end
      local matched = match(t, w)
      if filter == "not" then
        excludes = excludes + matched
      else
        resultset = resultset * matched
      end
      filter = nil
      ::nextw::
    end
    resultset = resultset - excludes
    subjects = Set:new(resultset)
    ::nextpart::
  end
  return resultset
end

function ElementNode:select(s) return select(self, s) end
ElementNode.mt.__call = select

return ElementNode