Module:TableTools

From WPRDC Wiki
Revision as of 07:44, 19 December 2013 by >Mr. Stradivarius (add isNan function, shallowClone function and removeDuplicates function, fix up valueIntersection function to work properly for NaNs)
Jump to navigation Jump to search

Documentation for this module may be created at Module:TableTools/doc

--[[
------------------------------------------------------------------------------------
--                               TableTools                                       --
--                                                                                --
-- This module includes a number of functions for dealing with Lua tables.        --
-- It is a meta-module, meant to be called from other Lua modules, and should     --
-- not be called directly from #invoke.                                           --
------------------------------------------------------------------------------------
--]]

local libraryUtil = require('libraryUtil')

local p = {}

-- Define often-used variables and functions.
local floor = math.floor
local infinity = math.huge
local checkType = libraryUtil.checkType

-- Define a unique value to represent NaN. This is because NaN cannot be used as a table key.
local nan = {}

--[[
------------------------------------------------------------------------------------
-- isPositiveInteger
--
-- This function returns true if the given value is a positive integer, and false
-- if not. Although it doesn't operate on tables, it is included here as it is
-- useful for determining whether a given table key is in the array part or the
-- hash part of a table.
------------------------------------------------------------------------------------
--]]
function p.isPositiveInteger(v)
	if type(v) == 'number' and v >= 1 and floor(v) == v and v < infinity then
		return true
	else
		return false
	end
end

--[[
------------------------------------------------------------------------------------
-- isNan
--
-- This function returns true if the given number is a NaN value, and false
-- if not. Although it doesn't operate on tables, it is included here as it is
-- useful for determining whether a value can be a valid table key. Lua will
-- generate an error if a NaN is used as a table key.
------------------------------------------------------------------------------------
--]]
function p.isNan(v)
	if type(v) == 'number' and tostring(v) == '-nan' then
		return true
	else
		return false
	end
end

--[[
------------------------------------------------------------------------------------
-- shallowClone
--
-- This returns a clone of a table. The value returned is a new table, but all
-- subtables and functions are shared. Metamethods are respected, but the returned
-- table will have no metatable of its own.
------------------------------------------------------------------------------------
--]]
function p.shallowClone(t)
	local ret = {}
	for k, v in pairs(t) do
		ret[k] = v
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- removeDuplicates
--
-- This removes duplicate values from an array. Non-positive-integer keys are
-- ignored. The earliest value is kept, and all subsequent duplicate values are
-- removed, but otherwise the array order is unchanged.
------------------------------------------------------------------------------------
--]]
function p.removeDuplicates(t)
	local isNan = p.isNan
	local ret, exists = {}, {}
	for i, v in ipairs(t) do
		if isNan(v) then
			-- NaNs can't be table keys, and they are also unique, so we don't need to check existence.
			ret[#ret + 1] = v
		else
			if not exists[v] then
				ret[#ret + 1] = v
				exists[v] = true
			end
		end	
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- union
--
-- This returns the union of the key/value pairs of n tables. If any of the tables
-- contain different values for the same table key, the table value is converted
-- to an array holding all of the different values.
------------------------------------------------------------------------------------
--]]
function p.union(...)
	local lim = select('#', ...) 
	if lim == 0 then
		error("no arguments passed to 'union'", 2)
	end
	local ret, trackArrays = {}, {}
	for i = 1, lim do
		local t = select(i, ...)
		checkType('union', i, t, 'table')
		for k, v in pairs(t) do
			local retKey = ret[k]
			if retKey == nil then
				ret[k] = v
			elseif retKey ~= v then
				if trackArrays[k] then
					local array = ret[k]
					local valExists
					for _, arrayVal in ipairs(array) do
						if arrayVal == v then
							valExists = true
							break
						end
					end
					if not valExists then
						array[#array + 1] = v
						ret[k] = array
					end
				else
					ret[k] = {ret[k], v}
					trackArrays[k] = true
				end
			end
		end
	end
	return ret
end				

--[[
------------------------------------------------------------------------------------
-- valueUnion
--
-- This returns the union of the values of n tables, as an array. For example, for
-- the tables {1, 3, 4, 5, foo = 7} and {2, bar = 3, 5, 6}, union will return
-- {1, 2, 3, 4, 5, 6, 7}.
------------------------------------------------------------------------------------
--]]
function p.valueUnion(...)
	local lim = select('#', ...) 
	if lim == 0 then
		error("no arguments passed to 'valueUnion'", 2)
	end
	local vals, ret = {}, {}
	for i = 1, lim do
		local t = select(i, ...)
		checkType('valueUnion', i, t, 'table')
		for k, v in pairs(t) do
			if type(v) == 'number' and tostring(v) == '-nan' then
				v = nan -- NaN cannot be a table key, so use a proxy variable.
			end
			vals[v] = true
		end
	end
	for val in pairs(vals) do
		if val == nan then
			-- This ensures that we output a NaN when we had one as input, although
			-- they may have been generated in a completely different way.
			val = 0/0 
		end
		ret[#ret + 1] = val
	end
	return ret
end	

--[[
------------------------------------------------------------------------------------
-- intersection
--
-- This returns the intersection of the key/value pairs of n tables. Both the key
-- and the value must match to be included in the resulting table.
------------------------------------------------------------------------------------
--]]
function p.intersection(...)
	local lim = select('#', ...) 
	if lim == 0 then
		error("no arguments passed to 'intersection'", 2)
	end
	local ret, track, pairCounts = {}, {}, {}
	for i = 1, lim do
		local t = select(i, ...)
		checkType('intersection', i, t, 'table')
		for k, v in pairs(t) do
			local trackVal = track[k]
			if trackVal == nil then
				track[k] = v
				pairCounts[k] = 1
			elseif trackVal == v then
				pairCounts[k] = pairCounts[k] + 1
			end
		end
	end
	for k, v in pairs(track) do
		if pairCounts[k] == lim then
			ret[k] = v
		end
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- valueIntersection
--
-- This returns the intersection of the values of n tables, as an array. For
-- example, for the tables {1, 3, 4, 5, foo = 7} and {2, bar = 3, 5, 6}, 
-- intersection will return {3, 5}.
------------------------------------------------------------------------------------
--]]
function p.valueIntersection(...)
	local lim = select('#', ...) 
	if lim < 2 then
		error(lim .. ' argument' .. (lim == 1 and '' or 's') .. " passed to 'intersection' (minimum is 2)", 2)
	end
	local isNan = p.isNan
	local vals, ret = {}, {}
	local isSameTable = true -- Tracks table equality.
	local tableTemp -- Used to store the table from the previous loop so that we can check table equality.
	for i = 1, lim do
		local t = select(i, ...)
		checkType('valueIntersection', i, t, 'table')
		if tableTemp and t ~= tableTemp then
			isSameTable = false
		end
		tableTemp = t
		for k, v in pairs(t) do
			-- NaNs are never equal to any other value, so they can't be in the intersection.
			-- Which is lucky, as they also can't be table keys.
			if not isNan(v) then
				local valCount = vals[v] or 0
				vals[v] = valCount + 1
			end
		end
	end
	if isSameTable then
		-- If all the tables are equal, then the intersection is that table (including NaNs).
		-- All we need to do is convert it to an array and remove duplicate values.
		for k, v in pairs(tableTemp) do
			ret[#ret + 1] = v
		end
		return p.removeDuplicates(ret)
	end
	for val, count in pairs(vals) do
		if count == lim then
			ret[#ret + 1] = val
		end
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- complement
--
-- This returns the relative complement of t1, t2, ..., in tn. The complement
-- is of key/value pairs. This is equivalent to all the key/value pairs that are in
-- tn but are not in t1, t2, ... tn-1.
------------------------------------------------------------------------------------
--]]
function p.complement(...)
	local lim = select('#', ...) 
	if lim == 0 then
		error("no arguments passed to 'complement' (minimum is two)", 2)
	elseif lim == 1 then
		error("only one argument passed to 'complement' (minimum is two)", 2)
	end
	--[[
	-- Now we know that we have at least two sets.
	-- First, get all the key/value pairs in tn. We can't simply make ret equal to tn,
	-- as that will affect the value of tn for the whole module.
	--]]
	local tn = select(lim, ...)
	checkType('complement', lim, tn, 'table')
	local ret = {}
	for k, v in pairs(tn) do
		ret[k] = v
	end
	-- Remove all the key/value pairs in t1, t2, ..., tn-1.
	for i = 1, lim - 1 do
		local t = select(i, ...)
		checkType('complement', i, t, 'table')
		for k, v in pairs(t) do
			if ret[k] == v then
				ret[k] = nil
			end
		end
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- numKeys
--
-- This takes a table and returns an array containing the numbers of any numerical
-- keys that have non-nil values, sorted in numerical order.
------------------------------------------------------------------------------------
--]]
function p.numKeys(t)
	checkType('numKeys', 1, t, 'table')
	local isPositiveInteger = p.isPositiveInteger
	local nums = {}
	for k, v in pairs(t) do
		if isPositiveInteger(k) then
			nums[#nums + 1] = k
		end
	end
	table.sort(nums)
	return nums
end

--[[
------------------------------------------------------------------------------------
-- affixNums
--
-- This takes a table and returns an array containing the numbers of keys with the
-- specified prefix and suffix. For example, for the table
-- {a1 = 'foo', a3 = 'bar', a6 = 'baz'} and the prefix "a", affixNums will
-- return {1, 3, 6}.
------------------------------------------------------------------------------------
--]]
function p.affixNums(t, prefix, suffix)
	checkType('affixNums', 1, t, 'table')
	checkType('affixNums', 2, prefix, 'string', true)
	checkType('affixNums', 3, suffix, 'string', true)
	prefix = prefix or ''
	suffix = suffix or ''
	local pattern = '^' .. prefix .. '([1-9]%d*)' .. suffix .. '$'
	local nums = {}
	for k, v in pairs(t) do
		if type(k) == 'string' then			
			local num = mw.ustring.match(k, pattern)
			if num then
				nums[#nums + 1] = tonumber(num)
			end
		end
	end
	table.sort(nums)
	return nums
end

--[[
------------------------------------------------------------------------------------
-- compressSparseArray
--
-- This takes an array with one or more nil values, and removes the nil values
-- while preserving the order, so that the array can be safely traversed with
-- ipairs.
------------------------------------------------------------------------------------
--]]
function p.compressSparseArray(t)
	checkType('compressSparseArray', 1, t, 'table')
	local ret = {}
	local nums = p.numKeys(t)
	for _, num in ipairs(nums) do
		ret[#ret + 1] = t[num]
	end
	return ret
end

--[[
------------------------------------------------------------------------------------
-- sparseIpairs
--
-- This is an iterator for sparse arrays. It can be used like ipairs, but can
-- handle nil values.
------------------------------------------------------------------------------------
--]]
function p.sparseIpairs(t)
	checkType('sparseIpairs', 1, t, 'table')
	local nums = p.numKeys(t)
	local i = 0
	local lim = #nums
	return function ()
		i = i + 1
		if i <= lim then
			local key = nums[i]
			return key, t[key]
		end
	end
end

--[[
------------------------------------------------------------------------------------
-- size
--
-- This returns the size of a key/value pair table. It will also work on arrays,
-- but for arrays it is more efficient to use the # operator.
------------------------------------------------------------------------------------
--]]
function p.size(t)
	checkType('size', 1, t, 'table')
	local i = 0
	for k in pairs(t) do
		i = i + 1
	end
	return i
end

return p