Content deleted Content added
m Updated TableTools dependency |
Implemented hash collision handling by adding a function hash table and, most importantly, by greatly improving the internal data representation with "_elements" and "_elements_obj". Improved valid element check. Converted "_get", "_set" and "_del" to normal methods. Optimized some methods by using "_get" instead of "has". |
||
Line 2:
local libraryUtil = require('libraryUtil')
local TableTools = require('Module:TableTools')
local warn = require('Module:Warning')
local basic_types = {boolean=1, string=1, number=1, ['function']=1}
local f_hashes = {} -- so that function elements can be properly compared and ordered in frozenset._hash
local f_hashes_mt = {
__index = function (t, key)
local h = tonumber('0x' .. mw.hash.hashValue('fnv1a32', tostring(os.time() + math.random())))
f_hashes[key] = h
return h
end,
__mode = 'k'
}
setmetatable(f_hashes, f_hashes_mt)
local frozenset, _frozenset = class('frozenset', {
__init = function (self, args)
local elements = {}
self._elements = elements
self._elements_obj = {}
if #args == 0 then -- for performance
return
elseif #args == 1 then
local arg = args[1]
if type(arg) == 'string' then
local c
for i = 1, mw.ustring.len(arg) do
elements[c] = c
end
return
elseif pcall(pairs, arg) or pcall(ipairs, arg) then
Line 26 ⟶ 43:
if TableTools.isArrayLike(args) then
for i, v in ipairs(args) do
if type(
error(("TypeError: invalid element #%d type (got %s,
end
self._set(
end
else
for k in pairs(args) do
if type(
error(("TypeError: invalid element type (got %s,
end
self._set(
end
end
end,
_get = function (self, elem)
if basic_types[type(elem)] then
return self._elements[elem]
else
local elements_obj = self._elements_obj
local h = elem.hash()
while elements_obj[h] ~= nil do
if elements_obj[h] == elem then
return elem
end
h = h + 1
end
end
end,
_set = function (self, elem)
if basic_types[type(elem)] then
self._elements[elem] = elem
else
local elements_obj = self._elements_obj
local h = elem.hash()
while elements_obj[h] ~= nil do
if elements_obj[h] == elem then
return
end
h = h + 1
end
elements_obj[h] = elem -- otherwise different objects with the same content would duplicate
end
end,
__pairs = function (self)
local elems = self._elements
local k, v
local function iterator()
if k == nil and elems == self._elements then
elems = self._elements_obj
end
return v -- nil at the very end
end
return iterator
end,
Line 60 ⟶ 111:
end,
-- "number" < "string", so numbers will be sorted before strings.
local type1, type2 = type(item1), type(item2)
if type1 ~= type2 then
return type1 < type2
elseif type1 == 'number' or type1 == 'string' then
return item1 < item2
elseif type1 == 'boolean' then
return tostring(item1) < tostring(item2)
else
local hash1, hash2
if type1 == 'function' then
hash1, hash2 = f_hashes[item1], f_hashes[item2]
if hash1 == hash2 then
warn(("HashWarning: function hash collision at %d"):format(hash1), 2)--what should be the level?
end
else
hash1, hash2 = item1.hash(), item2.hash()
if hash1 == hash2 then
warn(("HashWarning: object hash collision at %d"):format(hash1), 2)
end
end
return hash1 < hash2
end
end,
_hash = function (self)
if self.__hash ~= nil then
return self.__hash
end
-- frozensets with the same elements (meaning equal) may have a different order, so 'order' them before hashing
local ordered_elems = TableTools.keysToList(self, self._keySort, true)
-- convert elements to strings for table.concat
local elemType
for i, elem in ipairs(ordered_elems) do
elemType = type(elem)
if elemType == 'number' or elemType == 'boolean' then
ordered_elems[i] = tostring(elem)
elseif elemType == 'string' then
ordered_elems[i] = "'" .. elem .. "'"
elseif elemType == 'function' then
ordered_elems[i] = 'f' .. f_hashes[elem]
else
end
end
local str = '{' .. table.concat(
self.__hash = tonumber('0x' .. mw.hash.hashValue('fnv1a32', str))
return self.__hash
Line 98 ⟶ 167:
__tostring = function (self)
local string_elems = {}
local elemType
for elem in pairs(self) do
if elemType == 'string' then
string_elems[#string_elems+1] = "'" .. elem .. "'"
elseif elemType == 'function' then
string_elems[#string_elems+1] = 'f' .. f_hashes[elem]
else
string_elems[#string_elems+1] = tostring(elem)
end
end
local str = '{' .. table.concat(string_elems, ', ') .. '}'
return str
Line 110 ⟶ 185:
len = function (self)
return TableTools.size(self._elements) + TableTools.size(self._elements_obj)
end,
Line 116 ⟶ 191:
if isinstance(elem, 'set') then
elem = frozenset{elem}
elseif type(
error(("TypeError: invalid element type (got %s,
end
return self._get(
end,
Line 125 ⟶ 200:
libraryUtil.checkTypeMulti('isdisjoint', 1, other, {'set', 'frozenset'})
for elem in pairs(other) do
if self._get(
return false
end
Line 138 ⟶ 213:
__le = function (a, b)
for elem in pairs(a) do
if
return false
end
Line 160 ⟶ 235:
__add = function (a, b)
local
for elem in pairs(a) do
end
for elem in pairs(b) do
end
return
end,
Line 177 ⟶ 252:
__mul = function (a, b)
local
for elem in pairs(a) do
if b._get(
end
end
return
end,
Line 193 ⟶ 268:
__sub = function (a, b)
local
for elem in pairs(a) do
if
end
end
return
end,
Line 207 ⟶ 282:
__pow = function (a, b)
local
for elem in pairs(a) do
if
end
end
for elem in pairs(b) do
if
end
end
return
end,
copy = function (self)
return (self.__class{self}) -- to not leak the private instance
end,
Line 229 ⟶ 304:
end,
__staticmethods = {'
__protected = {'_get', '_set'}
})
Line 235 ⟶ 310:
local set = class('set', frozenset, {
_del = function (
if
else
local elements_obj = self._elements_obj
local h = elem.hash()
while elements_obj[h] ~= nil do
if elements_obj[h] == elem then
elements_obj[h] = nil
return
end
h = h + 1
end
end
end,
update = function (self, ...)
local others, other = {...}
for i = 1, select('#', ...) do
other = frozenset{others[i]}
for elem in pairs(other) do
self._set(
end
end
Line 254 ⟶ 338:
intersection_update = function (self, ...)
local others,
for i = 1, select('#', ...) do
for elem in pairs(self) do -- probably faster than iterating through (likely longer) "other"
if _other._get(elem) == nil then
self._del(
end
end
Line 266 ⟶ 350:
difference_update = function (self, ...)
local others,
for i = 1, select('#', ...) do
for elem in pairs(self) do
if
self._del(
end
end
Line 278 ⟶ 362:
symmetric_difference_update = function (self, other)
for elem in pairs(self) do
if
self._del(
end
end
for elem in pairs(
if
self._set(
end
end
Line 292 ⟶ 376:
add = function (self, elem)
if type(
error(("TypeError: invalid element type (got %s,
end
self._set(
end,
Line 301 ⟶ 385:
if isinstance(elem, 'set') then
elem = frozenset{elem}
elseif type(
error(("TypeError: invalid element type (got %s,
end
if
error(("KeyError: %s"):format(tostring(elem)), 2)
end
self._del(
end,
Line 313 ⟶ 397:
if isinstance(elem, 'set') then
elem = frozenset{elem}
elseif type(
error(("TypeError: invalid element type (got %s,
end
self._del(
end,
Line 322 ⟶ 406:
local k, v = next(self._elements)
if k == nil then
k, v = next(self._elements_obj)
if k == nil then
error("KeyError: pop from an empty set", 2)
end
end
self._del(v)
return v
end,
clear = function (self)
self._elements = {}
self._elements_obj = {}
end,
__protected = {'_del'}
})
|