Modul:table

Ushbu modul Lua jadvallari bilan ishlash funksiyalarini taqdim etadi. Ularning barchasi, ikkita yordamchi funksiyadan tashqari, birinchi dalil sifatida jadvalni oladi.

Ayrim funksiyalar Module:array tomonidan yaratilgan massivlarda usullar sifatida mavjud.

Funktsiyalari nimaga koʻra:

• Yangi jadval yarating:
  • shallowcopy, deepcopy, removeDuplicates, numKeys, compressSparseArray, keysToList, reverse, invert, listToSet
• Massiv yarating:
  • removeDuplicates, numKeys, compressSparseArray, keysToList, reverse
• Jadval haqida ma'lumotni qaytarish:
  • size, length, contains, keyFor, isArray, deepEquals
• Jadvalni massiv sifatida ko'rib chiqing (ya'ni, jadvalning massiv qismidagi qiymatlar bilan ishlang: 1 dan boshlanadigan ketma-ket butun sonlar bilan indekslangan qiymatlar):
  • removeDuplicates, length, contains, serialCommaJoin, reverseIpairs, reverse, invert, listToSet, isArray
• Jadvalni siyrak massiv sifatida ko'rib chiqing (ya'ni, ketma-ket bo'lmagan butun sonlar bilan indekslangan qiymatlar ustida ishlang):
  • numKeys, maxIndex, compressSparseArray, sparseConcat, sparseIpairs
• Iterator yarating:
  • sparseIpairs, sortedPairs, reverseIpairs
• Boshqa:
  • sparseConcat, serialCommaJoin, reverseConcat

Asl versiya Commonsdagi Module:TableTools orqali Vikipediyadagi Module:TableTools nusxasi edi, biroq oʻshandan beri yangi funksiyalar qoʻshildi.

Batafsil hujjatlar

export.isPositiveInteger

function export.isPositiveInteger(n)

Return 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.

export.shallowcopy

function export.shallowcopy(orig, raw)

Return a clone of an object. If the object is a table, the value returned is a new table, but all subtables and functions are shared. Metamethods are respected unless the raw flag is set, but the returned table will have no metatable of its own.

export.deepcopy

function export.deepcopy(orig, metatableFlag, keepLoadedData)

Recursive deep copy function. Preserves copied identities of subtables. A more powerful version of mw.clone, with customizable options.

• By default, metatables are copied, except for data loaded via mw.loadData (see below). If metatableFlag is set to "none", the copy will not have any metatables at all. Conversely, if metatableFlag is set to "keep", then the cloned table (and all its members) will have the exact same metatable as their original version.
• If keepLoadedData is true, then any data loaded via mw.loadData will not be copied, and the original will be used instead. This is useful in iterative contexts where it is necessary to copy data being destructively modified, because objects loaded via mw.loadData are immutable.
• Notes:
  1. Protected metatables will not be copied (i.e. those hidden behind a metatable metamethod), as they are not accessible by Lua's design. Instead, the output of the metatable method will be used instead.
  2. When iterating over the table, the pairs metamethod is ignored, since this can prevent the table from being properly cloned.
  3. Data loaded via mw.loadData is a special case in two ways: the metatable is stripped, because otherwise the cloned table throws errors when accessed; in addition, the pairs metamethod is used, since otherwise the cloned table would be empty.

export.append

function export.append(...)

Append any number of tables together and returns the result. Compare the Lisp expression (append list1 list2 ...).

export.extendList

function export.extendList(list, new_items, options)

Extend an existing list by a new list, modifying the existing list in-place. Compare the Python expression list.extend(new_items).

options is an optional table of additional options to control the behavior of the operation. The following options are recognized:

• insertIfNot: Use export.insertIfNot() instead of table.insert(), which ensures that duplicate items do not get inserted (at the cost of an O((M+N)*N) operation, where M = #list and N = #new_items).
• key: As in insertIfNot(). Ignored otherwise.
• pos: As in insertIfNot(). Ignored otherwise.

export.removeDuplicates

function export.removeDuplicates(t)

Remove 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.

export.numKeys

function export.numKeys(t)

Given a table, return an array containing all positive integer keys, sorted in numerical order.

export.maxIndex

function export.maxIndex(t)

Return the highest positive integer index of a table or array that possibly has holes in it, or otherwise 0 if no positive integer keys are found. Note that this differs from table.maxn, which returns the highest positive numerical index, even if it is not an integer.

export.compressSparseArray

function export.compressSparseArray(t)

This takes a list with one or more nil values, and removes the nil values while preserving the order, so that the list can be safely traversed with ipairs.

export.rawPairs

function export.rawPairs(t)

An iterator which works like pairs, but ignores any __pairs metamethod.

export.rawIpairs

function export.rawIpairs(t)

An iterator which works like ipairs, but ignores any __ipairs metamethod.

export.indexPairs

function export.indexPairs(t, raw)

An iterator which works like pairs, except that it also respects the index metamethod. This works by iterating over the input table with pairs, followed by the table at its index metamethod (if any). This is then repeated for that table's index table and so on, with any repeated keys being skipped over, until there are no more tables, or a table repeats (so as to prevent an infinite loop). If index is a function, however, then it is ignored, since there is no way to iterate over its return values. A pairs metamethod will be respected for any given table instead of iterating over it directly, but these will be ignored if the raw flag is set.

Note: this function can be used as a pairs metamethod. In such cases, it does not call itself, since this would cause an infinite loop, so it treats the relevant table as having no pairs metamethod. Other pairs metamethods on subsequent tables will still be respected.

export.indexIpairs

function export.indexIpairs(t, raw)

An iterator which works like ipairs, except that it also respects the index metamethod. This works by looking up values in the table, iterating integers from key 1 until no value is found. An ipairs metamethod for the input table will be respected, but it will be ignored if the raw flag is set.

Note: this function can be used as an ipairs metamethod. In such cases, it does not call itself, since this would cause an infinite loop, so it treats the input table as having no ipairs metamethod.

export.sparseIpairs

function export.sparseIpairs(t)

This is an iterator for sparse arrays. It can be used like ipairs, but can handle nil values.

export.size

function export.size(t, raw)

This returns the size of a key/value pair table. If raw is set, then metamethods will be ignored, giving the true table size.

For arrays, it is faster to use export.length.

export.length

function export.length(t, raw)

This returns the length of a table, or the first integer key n counting from 1 such that t[n + 1] is nil. It is a more reliable form of the operator #, which can become unpredictable under certain circumstances due to the implementation of tables under the hood in Lua, and therefore should not be used when dealing with arbitrary tables. # also does not use metamethods, so will return the wrong value in cases where it is desirable to take these into account (e.g. data loaded via mw.loadData). If raw is set, then metamethods will be ignored, giving the true table length.

For arrays, this function is faster than export.size.

export.deepEquals

function export.deepEquals(a, b, include_mt, raw)

Recursively compare two values that may be tables, and returns true if all key-value pairs are structurally equivalent. Note that this handles arbitrary nesting of subtables (including recursive nesting) to any depth, for keys as well as values.

If include_mt is true, then metatables are also compared. If raw is true, then metamethods are not used during the comparison.

export.getNested

function export.getNested(t, ...)

Given a table and an arbitrary number of keys, will successively access subtables using each key in turn, returning the value at the final key. For example, if t is {[1] = {[2] = {[3] = "foo"}}}, export.getNested(t, 1, 2, 3) will return "foo".

If no subtable exists for a given key value, returns nil, but will throw an error if a non-table is found at an intermediary key.

export.setNested

function export.setNested(t, ...)

Given a table, value and an arbitrary number of keys, will successively access subtables using each key in turn, and sets the value at the final key. For example, if t is {}, export.setNested(t, "foo", 1, 2, 3) will modify t to {[1] = {[2] = {[3] = "foo"} } }.

If no subtable exists for a given key value, one will be created, but the function will throw an error if a non-table value is found at an intermediary key.

Note: the parameter order (table, value, keys) differs from functions like rawset, because the number of keys can be arbitrary. This is to avoid situations where an additional argument must be appended to arbitrary lists of variables, which can be awkward and error-prone: for example, when handling variable arguments (...) or function return values.

export.contains

function export.contains(list, x, options)

Given a list and a value to be found, return true if the value is in the array portion of the list. Comparison is by value, using deepEquals.

export.tableContains

function export.tableContains(t, x)

Given a general table and a value to be found, return true if the value is in either the array or hashmap portion of the table. Comparison is by value, using deepEquals.

export.insertIfNot

function export.insertIfNot(list, new_item, options)

Given a list and a new_item to be inserted, append the value to the end of the list if not already present (or insert at an arbitrary position, if options.pos is given; see below). Comparison is by value, using deepEquals.

options is an optional table of additional options to control the behavior of the operation. The following options are recognized:

• pos: Position at which insertion happens (i.e. before the existing item at position pos).
• key: Function of one argument to return a comparison key, as with deepEquals. The key function is applied to both item and the existing item in list to compare against, and the comparison is done against the results. This is useful when inserting a complex structure into an existing list while avoiding duplicates.
• combine: Function of three arguments (the existing item, the new item and the position, respectively) to combine an existing item with new_item, when new_item is found in list. If unspecified, the existing item is left alone.

Return false if entry already found, true if inserted.

For compatibility, pos can be specified directly as the third argument in place of options, but this is not recommended for new code.

NOTE: This function is O(N) in the size of the existing list. If you use this function in a loop to insert several items, you will get O(M*(M+N)) behavior, effectively O((M+N)^2). Thus it is not recommended to use this unless you are sure the total number of items will be small. (An alternative for large lists is to insert all the items without checking for duplicates, and use removeDuplicates() at the end.)

export.keyFor

function export.keyFor(t, x)

Finds key for specified value in a given table. Roughly equivalent to reversing the key-value pairs in the table:

• reversed_table = { [value1] = key1, [value2] = key2, ... } and then returning reversed_table[value]. Comparison is by value, using deepEquals.

Only reliable if there is just one key with the specified value. Otherwise, the function returns the first key found, and the output is unpredictable.

export.keysToList

function export.keysToList(t, keySort)

Return a list of the keys in a table, sorted using either the default table.sort function or a custom keySort function.

If there are only numerical keys, export.numKeys is probably faster.

export.sortedPairs

function export.sortedPairs(t, keySort)

Iterates through a table, with the keys sorted using the keysToList function.

If there are only numerical keys, export.sparseIpairs is probably faster.

export.reverseIpairs

function export.reverseIpairs(t)

export.reduce

function export.reduce(t, func, i, j, step)

Given an array list and function func, iterate through the array applying func(r, k, v), and returning the result, where r is the value calculated so far, k is an index, and v is the value at index k. For example, reduce(array, function(a, b) return a + b end) will return the sum of array.

Optional arguments:

• i: start index; negative values count from the end of the array
• j: end index; negative values count from the end of the array
• step: step increment These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:

1. No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
2. step=-1 results in backward iteration from the end to the start in steps of 1.
3. i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
4. j=-3 results in forward iteration from the start to the 3rd last index.
5. j=-3, step=-1 results in backward iteration from the end to the 3rd last index. Note: directionality generally only matters for reduce, but values of step > 1 (or step < -1) still affect the return value of apply.

export.apply

function export.apply(t, func, i, j, step)

Given an array list and function func, iterate through the array applying func(k, v) (where k is an index, and v is the value at index k), and return an array of the resulting values. For example, apply(array, function(a) return 2*a end) will return an array where each member of array has been doubled.

Optional arguments:

• i: start index; negative values count from the end of the array
• j: end index; negative values count from the end of the array
• step: step increment These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:

1. No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
2. step=-1 results in backward iteration from the end to the start in steps of 1.
3. i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
4. j=-3 results in forward iteration from the start to the 3rd last index.
5. j=-3, step=-1 results in backward iteration from the end to the 3rd last index. Note: directionality makes the most difference for reduce, but values of step > 1 (or step < -1) still affect the return value of apply.

export.all

function export.all(t, func, i, j, step)

Given an array list and function func, iterate through the array applying func(k, v) (where k is an index, and v is the value at index k), and returning whether the function is true for all iterations.

Optional arguments:

• i: start index; negative values count from the end of the array
• j: end index; negative values count from the end of the array
• step: step increment These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:

1. No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
2. step=-1 results in backward iteration from the end to the start in steps of 1.
3. i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
4. j=-3 results in forward iteration from the start to the 3rd last index.
5. j=-3, step=-1 results in backward iteration from the end to the 3rd last index.

export.any

function export.any(t, func, i, j, step)

Given an array list and function func, iterate through the array applying func(k, v) (where k is an index, and v is the value at index k), and returning whether the function is true for at least one iteration.

Optional arguments:

• i: start index; negative values count from the end of the array
• j: end index; negative values count from the end of the array
• step: step increment These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:

1. No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
2. step=-1 results in backward iteration from the end to the start in steps of 1.
3. i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
4. j=-3 results in forward iteration from the start to the 3rd last index.
5. j=-3, step=-1 results in backward iteration from the end to the 3rd last index.

export.serialCommaJoin

function export.serialCommaJoin(seq, options)

Joins an array with serial comma and serial conjunction, normally "and". An improvement on mw.text.listToText, which doesn't properly handle serial commas.

Options:

• conj: Conjunction to use; defaults to "and".
• punc: Punctuation to use; default to ",".
• dontTag: Don't tag the serial comma and serial "and". For error messages, in which HTML cannot be used.

export.concat

function export.concat(t, sep, i, j)

A function which works like table.concat, but respects any __index metamethod. This is useful for data loaded via mw.loadData.

export.sparseConcat

function export.sparseConcat(t, sep, i, j)

Concatenate all values in the table that are indexed by a number, in order.

• sparseConcat{ a, nil, c, d } => "acd"
• sparseConcat{ nil, b, c, d } => "bcd"

export.reverse

function export.reverse(t)

Values of numeric keys in array portion of table are reversed: { "a", "b", "c" } -> { "c", "b", "a" }

export.reverseConcat

function export.reverseConcat(t, sep, i, j)

export.invert

function export.invert(list)

Invert a list. For example, invert({ "a", "b", "c" }) -> { a = 1, b = 2, c = 3 }

export.listToSet

function export.listToSet(list, value, ...)

Convert list (a table with a list of values) into a set (a table where those values are keys instead). This is a useful way to create a fast lookup table, since looking up a table key is much, much faster than iterating over the whole list to see if it contains a given value.

By default, each item is given the value true. If the optional parameter value is a function or functor, then it is called as an iterator, with the list index as the first argument, the item as the second (which will be used as the key), plus any additional arguments passed to listToSet; the returned value is used as the value for that list item. If value is anything else, then it is used as the fixed value for every item.

export.isArray

function export.isArray(t)

Return true if all keys in the table are consecutive integers starting at 1.

export.alias

function export.alias(t, k, aliases)

Add a list of aliases for a given key to a table. The aliases must be given as a table.

---

--[[
------------------------------------------------------------------------------------
--                      table (formerly 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 export = {}

local collation_module = "Module:collation"
local debug_track_module = "Module:debug/track"
local function_module = "Module:fun"

local table = table

local concat = table.concat
local contains -- defined as export.contains
local deep_copy -- defined as export.deepcopy
local deep_equals -- defined as export.deepEquals
local format = string.format
local getmetatable = getmetatable
local insert = table.insert
local insert_if_not -- defined as export.insertIfNot
local invert -- defined as export.invert
local ipairs = ipairs
local ipairs_default_iter = ipairs{export}
local is_positive_integer -- defined as export.isPositiveInteger
local keys_to_list -- defined as export.keysToList
local next = next
local num_keys -- defined as export.numKeys
local pairs = pairs
local pcall = pcall
local raw_pairs -- defined as export.rawPairs
local rawequal = rawequal
local rawget = rawget
local require = require
local select = select
local setmetatable = setmetatable
local sort = table.sort
local sparse_ipairs -- defined as export.sparseIpairs
local table_len -- defined as export.length
local table_reverse -- defined as export.reverse
local type = type

local infinity = math.huge

--[==[
Loaders for functions in other modules, which overwrite themselves with the target function when called. This ensures modules are only loaded when needed, retains the speed/convenience of locally-declared pre-loaded functions, and has no overhead after the first call, since the target functions are called directly in any subsequent calls.]==]
	local function debug_track(...)
		debug_track = require(debug_track_module)
		return debug_track(...)
	end
	
	local function is_callable(...)
		is_callable = require(function_module).is_callable
		return is_callable(...)
	end
	
	local function string_sort(...)
		string_sort = require(collation_module).string_sort
		return string_sort(...)
	end

--[==[
Return 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 export.isPositiveInteger(n)
	return type(n) == "number" and n >= 1 and n % 1 == 0 and n < infinity
end
is_positive_integer = export.isPositiveInteger

--[==[
Return a clone of an object. If the object is a table, the value returned is a new table, but all subtables and functions are shared. Metamethods are respected unless the `raw` flag is set, but the returned table will have no metatable of its own.]==]
function export.shallowcopy(orig, raw)
	if type(orig) ~= "table" then
		return orig
	end
	local copy, iter, state, init = {}
	if raw then
		iter, state = next, orig
	else
		iter, state, init = pairs(orig)
		-- Track instances of data loaded via mw.loadData being copied, which is very inefficient and usually unnecessary.
		local mt = getmetatable(orig)
		if mt and type(mt) == "table" and rawget(mt, "mw_loadData") == true then
			debug_track("table/shallowcopy/loaded data")
		end
	end
	for k, v in iter, state, init do
		copy[k] = v
	end
	return copy
end

do
	local function make_copy(orig, memo, mt_flag, keep_loaded_data, tracked)
		if type(orig) ~= "table" then
			return orig
		end
		local memoized = memo[orig]
		if memoized ~= nil then
			return memoized
		end
		local mt, iter, state, init = getmetatable(orig)
		-- `mt` could be a non-table if `__metatable` has been used, but discard it in such cases.
		if not (mt and type(mt) == "table") then
			mt, iter, state, init = nil, next, orig, nil
		-- Data loaded via `mw.loadData`.
		elseif rawget(mt, "mw_loadData") == true then
			if keep_loaded_data then
				memo[orig] = orig
				return orig
			-- Track instances of such data being copied, which is very inefficient and usually unnecessary.
			elseif not tracked then
				debug_track("table/deepcopy/loaded data")
				tracked = true
			end
			-- Discard the metatable, and use the `__pairs` metamethod.
			mt, iter, state, init = nil, pairs(orig)
		-- Otherwise, use `mt`.
		else
			iter, state, init = next, orig, nil
		end
		local copy = {}
		memo[orig] = copy
		for k, v in iter, state, init do
			copy[make_copy(k, memo, mt_flag, keep_loaded_data, tracked)] = make_copy(v, memo, mt_flag, keep_loaded_data, tracked)
		end
		if mt == nil or mt_flag == "none" then
			return copy
		elseif mt_flag ~= "keep" then
			mt = make_copy(mt, memo, mt_flag, keep_loaded_data, tracked)
		end
		return setmetatable(copy, mt)
	end

	--[==[
	Recursive deep copy function. Preserves copied identities of subtables.
	A more powerful version of {mw.clone}, with customizable options.
	* By default, metatables are copied, except for data loaded via mw.loadData (see below). If `metatableFlag` is set to "none", the copy will not have any metatables at all. Conversely, if `metatableFlag` is set to "keep", then the cloned table (and all its members) will have the exact same metatable as their original version.
	* If `keepLoadedData` is true, then any data loaded via {mw.loadData} will not be copied, and the original will be used instead. This is useful in iterative contexts where it is necessary to copy data being destructively modified, because objects loaded via mw.loadData are immutable.
	* Notes:
	*# Protected metatables will not be copied (i.e. those hidden behind a __metatable metamethod), as they are not
	   accessible by Lua's design. Instead, the output of the __metatable method will be used instead.
	*# When iterating over the table, the __pairs metamethod is ignored, since this can prevent the table from being properly cloned.
	*# Data loaded via mw.loadData is a special case in two ways: the metatable is stripped, because otherwise the cloned table throws errors when accessed; in addition, the __pairs metamethod is used, since otherwise the cloned table would be empty.]==]
	function export.deepcopy(orig, metatableFlag, keepLoadedData)
		return make_copy(orig, {}, metatableFlag, keepLoadedData)
	end
	deep_copy = export.deepcopy
end

--[==[
Append any number of tables together and returns the result. Compare the Lisp expression {(append list1 list2 ...)}.]==]
function export.append(...)
	local args, list, n = {...}, {}, 0
	for i = 1, select("#", ...) do
		local t, j = args[i], 0
		while true do
			j = j + 1
			local v = t[j]
			if v == nil then
				break
			end
			n = n + 1
			list[n] = v
		end
	end
	return list
end

--[==[
Extend an existing list by a new list, modifying the existing list in-place. Compare the Python expression
{list.extend(new_items)}.

`options` is an optional table of additional options to control the behavior of the operation. The following options are
recognized:
* `insertIfNot`: Use {export.insertIfNot()} instead of {table.insert()}, which ensures that duplicate items do not get
  inserted (at the cost of an O((M+N)*N) operation, where M = #list and N = #new_items).
* `key`: As in {insertIfNot()}. Ignored otherwise.
* `pos`: As in {insertIfNot()}. Ignored otherwise.]==]
function export.extendList(list, new_items, options)
	local i, insert_if_not_option = 0, options and options.insertIfNot
	while true do
		i = i + 1
		local item = new_items[i]
		if item == nil then
			return
		elseif insert_if_not_option then
			insert_if_not(list, item, options)
		else
			insert(list, item)
		end
	end
end

do
	local pos_nan, neg_nan
	
	--[==[
	Remove 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 export.removeDuplicates(t)
		local list, seen, i, n = {}, {}, 0, 0
		while true do
			i = i + 1
			local v = t[i]
			if v == nil then
				return list
			end
			local memo_key
			if v == v then
				memo_key = v
			-- NaN
			elseif format("%f", v) == "nan" then
				if not pos_nan then
					pos_nan = {}
				end
				memo_key = pos_nan
			-- -NaN
			else
				if not neg_nan then
					neg_nan = {}
				end
				memo_key = neg_nan
			end
			if not seen[memo_key] then
				n = n + 1
				list[n], seen[memo_key] = v, true
			end
		end
		return list
	end
end

--[==[
Given a table, return an array containing all positive integer keys, sorted in numerical order.]==]
function export.numKeys(t)
	local nums, i = {}, 0
	for k in pairs(t) do
		if is_positive_integer(k) then
			i = i + 1
			nums[i] = k
		end
	end
	sort(nums)
	return nums
end
num_keys = export.numKeys

--[==[
Return the highest positive integer index of a table or array that possibly has holes in it, or otherwise 0 if no positive integer keys are found. Note that this differs from `table.maxn`, which returns the highest positive numerical index, even if it is not an integer.]==]
function export.maxIndex(t)
	local max = 0
	for k in pairs(t) do
		if is_positive_integer(k) and k > max then
			max = k
		end
	end
	return max
end

--[==[
This takes a list with one or more nil values, and removes the nil values while preserving the order, so that the list can be safely traversed with ipairs.]==]
function export.compressSparseArray(t)
	local list, keys, i = {}, num_keys(t), 0
	while true do
		i = i + 1
		local k = keys[i]
		if k == nil then
			return list
		end
		list[i] = t[k]
	end
end

--[==[
An iterator which works like `pairs`, but ignores any `__pairs` metamethod.]==]
function export.rawPairs(t)
	return next, t, nil
end
raw_pairs = export.rawPairs

--[==[
An iterator which works like `ipairs`, but ignores any `__ipairs` metamethod.]==]
function export.rawIpairs(t)
	return ipairs_default_iter, t, 0
end

do
	local current
	--[==[
	An iterator which works like `pairs`, except that it also respects the `__index` metamethod. This works by iterating over the input table with `pairs`, followed by the table at its `__index` metamethod (if any). This is then repeated for that table's `__index` table and so on, with any repeated keys being skipped over, until there are no more tables, or a table repeats (so as to prevent an infinite loop). If `__index` is a function, however, then it is ignored, since there is no way to iterate over its return values.
	
	A `__pairs` metamethod will be respected for any given table instead of iterating over it directly, but these will be ignored if the `raw` flag is set.

	Note: this function can be used as a `__pairs` metamethod. In such cases, it does not call itself, since this would cause an infinite loop, so it treats the relevant table as having no `__pairs` metamethod. Other `__pairs` metamethods on subsequent tables will still be respected.]==]
	function export.indexPairs(t, raw)
		-- If there's no metatable, result is identical to `pairs`.
		-- To prevent infinite loops, act like `pairs` if `current` is set with `t`, which means this function is being used as a __pairs metamethod.
		if current and current[t] or getmetatable(t) == nil then
			return next, t, nil
		end
		
		-- `seen_k` memoizes keys, as they should never repeat; `seen_t` memoizes tables iterated over.
		local seen_k, seen_t, iterate, iter, state = {}, {}
		
		local function catch_values(k, ...)
			if k ~= nil then
				-- If a repeated key is found, skip and iterate again.
				if seen_k[k] then
					return catch_values(iter(state, k))
				end
				seen_k[k] = true
				return k, ...
			end
			-- If there's an __index metamethod, iterate over it iff it's a table not already seen before. Otherwise, return.
			local mt = getmetatable(t)
			-- `mt` might not be a table if __metatable is used.
			if mt == nil or type(mt) ~= "table" then
				return nil
			end
			t = rawget(mt, "__index")
			if t == nil or type(t) ~= "table" or seen_t[t] then
				return nil
			end
			seen_t[t], iter = true, nil
			return iterate()
		end
		
		function iterate(_, k)
			if iter ~= nil then
				-- Metamethods can return an arbitrary number of values, so catch them to avoid creating a table.
				return catch_values(iter(state, k))
			end
			-- If `raw` is set, use `next`.
			if raw then
				iter, state, k = next, t, nil
			-- Otherwise, call `pairs`, setting `current` with `t` so that export.indexPairs knows to return `next` if it's being used as a metamethod, as this prevents infinite loops. `t` is then unset, so that `current` doesn't get polluted if the loop breaks early.
			else
				if not current then
					current = {}
				end
				current[t] = true
				-- Use `pcall`, so that `t` can always be unset from `current`.
				local success
				success, iter, state, k = pcall(pairs, t)
				current[t] = nil
				-- If there was an error, raise it.
				if not success then
					error(iter)
				end
			end
			return catch_values(iter(state, k))
		end
		
		return iterate
	end
end

do
	local current
	
	local function ipairs_func(t, i)
		i = i + 1
		local v = t[i]
		if v ~= nil then
			return i, v
		end
	end
	
	--[==[
	An iterator which works like `ipairs`, except that it also respects the `__index` metamethod. This works by looking up values in the table, iterating integers from key `1` until no value is found.
	
	An `__ipairs` metamethod for the input table will be respected, but it will be ignored if the `raw` flag is set.

	Note: this function can be used as an `__ipairs` metamethod. In such cases, it does not call itself, since this would cause an infinite loop, so it treats the input table as having no `__ipairs` metamethod.]==]
	function export.indexIpairs(t, raw)
		-- If there's no metatable, result is identical to `ipairs`.
		if getmetatable(t) == nil then
			if not current then
				current = {}
			end
			return ipairs_default_iter, t, 0
		-- To prevent infinite loops, don't check for a metamethod if `current` is set with `t`, which means this function is being used as an __ipairs metamethod. Also if `raw` is set.
		elseif raw or current and current[t] then
			return ipairs_func, t, 0
		elseif not current then
			current = {}
		end
		current[t] = true
		-- Use `pcall`, so that `t` can always be unset from `current`.
		local success, iter, state, init = pcall(ipairs, t)
		current[t] = nil
		-- If there was an error, raise it.
		if not success then
			error(iter)
		-- If `iter` is just the default `ipairs` iterator, then use `ipairs_func`. There could still be an __ipairs metamethod, but it's safe to use `ipairs_func` on whatever `state` and `init` values were returned.
		elseif iter == ipairs_default_iter then
			iter = ipairs_func
		end
		return iter, state, init
	end
end

--[==[
This is an iterator for sparse arrays. It can be used like ipairs, but can handle nil values.]==]
function export.sparseIpairs(t)
	local keys, i = num_keys(t), 0
	return function()
		i = i + 1
		local k = keys[i]
		if k then
			return k, t[k]
		end
	end
end
sparse_ipairs = export.sparseIpairs

--[==[
This returns the size of a key/value pair table. If `raw` is set, then metamethods will be ignored, giving the true table size.

For arrays, it is faster to use `export.length`.]==]
function export.size(t, raw)
	local i, iter, state, init = 0
	if raw then
		iter, state, init = next, t, nil
	else
		iter, state, init = pairs(t)
	end
	for _ in iter, state, init do
		i = i + 1
	end
	return i
end

--[==[
This returns the length of a table, or the first integer key n counting from 1 such that t[n + 1] is nil. It is a more reliable form of the operator `#`, which can become unpredictable under certain circumstances due to the implementation of tables under the hood in Lua, and therefore should not be used when dealing with arbitrary tables. `#` also does not use metamethods, so will return the wrong value in cases where it is desirable to take these into account (e.g. data loaded via `mw.loadData`). If `raw` is set, then metamethods will be ignored, giving the true table length.

For arrays, this function is faster than `export.size`.]==]
function export.length(t, raw)
	local n = 0
	if raw then
		for i in ipairs_default_iter, t, 0 do
			n = i
		end
		return n
	end
	repeat
		n = n + 1
	until t[n] == nil
	return n - 1
end
table_len = export.length

do
	local function is_equivalent(a, b, memo, include_mt, pairs_func)
		-- Raw equality check.
		if rawequal(a, b) then
			return true
		-- If not equal, a and b can only be equivalent if they're both tables.
		elseif not (type(a) == "table" and type(b) == "table") then
			return false
		end
		-- If `a` and `b` have been compared before, return the memoized result. This will usually be true, since failures normally fail the whole check outright, but match failures can occur during the laborious check without this happening, so it could be false.
		local memo_a = memo[a]
		if memo_a then
			local result = memo_a[b]
			if result ~= nil then
				return result
			end
			-- To avoid recursive references causing infinite loops, assume the tables currently being compared are equivalent by memoizing them as true; this will be corrected to false if there's a match failure.
			memo_a[b] = true
		else
			memo_a = {[b] = true}
			memo[a] = memo_a
		end
		-- Don't bother checking `memo_b` for `a`, since if `a` and `b` had been compared before, then `b` would be in `memo_a`.
		local memo_b = memo[b]
		if memo_b then
			memo_b[a] = true
		else
			memo_b = {[a] = true}
			memo[b] = memo_b
		end
		-- If `include_mt` is set, check the metatables are equivalent.
		if include_mt and not is_equivalent(getmetatable(a), getmetatable(b), memo, true, pairs_func) then
			memo_a[b], memo_b[a] = false, false
			return false
		end
		-- Copy all key/values pairs in `b` to `remaining_b`, and count the size: this uses `pairs_func`, which will also be used to iterate over `a`, ensuring that `a` and `b` are iterated over in the same way. This is necessary to ensure that `export.deepEquals(a, b)` and `export.deepEquals(b, a)` always give the same result. Simply iterating over `a` while accessing keys in `b` for comparison would ignore any `__pairs` metamethod that `b` has, which could cause non-symmetrical outputs if `__pairs` returns more or less than the complete set of key/value pairs accessible via `__index`, so using `pairs_func` for both `a` and `b` prevents this.
		-- TODO: handle exotic `__pairs` methods which return the same key multiple times with different values.
		local remaining_b, size_b = {}, 0
		for k_b, v_b in pairs_func(b) do
			remaining_b[k_b], size_b = v_b, size_b + 1
		end
		-- Fast check: iterate over the keys in `a`, checking if an equivalent value exists at the same key in `remaining_b`. As matches are found, key/value pairs are removed from `remaining_b`. If any keys in `a` or `remaining_b` are tables, the fast check will only work if the exact same object exists as a key in the other table. Any others from `a` that don't match anything in `remaining_b` are added to `remaining_a`, while those in `remaining_b` that weren't found will still remain once the loop ends. `remaining_a` and `remaining_b` are then compared at the end with the laborious check.
		local size_a, remaining_a = 0
		for k, v_a in pairs_func(a) do
			local v_b = remaining_b[k]
			-- If `k` isn't in `remaining_b`, `a` and `b` can't be equivalent unless it's a table.
			if v_b == nil then
				if type(k) ~= "table" then
					memo_a[b], memo_b[a] = false, false
					return false
				-- Otherwise, add the `k`/`v_a` pair to `remaining_a` for the laborious check.
				elseif not remaining_a then
					remaining_a = {}
				end
				remaining_a[k], size_a = v_a, size_a + 1
			-- Otherwise, if `k` exists in `a` and `remaining_b`, `v_a` and `v_b` must be equivalent for there to be a match.
			elseif is_equivalent(v_a, v_b, memo, include_mt, pairs_func) then
				remaining_b[k], size_b = nil, size_b - 1
			else
				memo_a[b], memo_b[a] = false, false
				return false
			end
		end
		-- Must be the same number of remaining keys in each table.
		if size_a ~= size_b then
			memo_a[b], memo_b[a] = false, false
			return false
		-- If the size is 0, there's nothing left to check.
		elseif size_a == 0 then
			return true
		end
		-- Laborious check: since it's not possible to use table lookups, check each key/value pair in `remaining_a` against every key/value pair in `remaining_b` until a match is found, removing the matching key/value pair from `remaining_b` each time, to ensure one-to-one correspondence.
		for k_a, v_a in next, remaining_a do
			local success
			for k_b, v_b in next, remaining_b do
				-- Keys/value pairs must be equivalent in order to match.
				if (
					-- Check values first for speed, since they might not be tables.
					is_equivalent(v_a, v_b, memo, include_mt, pairs_func) and
					is_equivalent(k_a, k_b, memo, include_mt, pairs_func)
				) then
					-- Remove matched key from `remaining_b`, and break the inner loop.
					success, remaining_b[k_b] = true, nil
					break
				end
			end
			-- Fail if `remaining_b` runs out of keys, as the `k_a`/`v_a` pair still hasn't matched.
			if not success then
				memo_a[b], memo_b[a] = false, false
				return false
			end
		end
		-- If every key/value pair in `remaining_a` matched with one in `remaining_b`, `a` and `b` must be equivalent. Note that `remaining_b` will now be empty, since the laborious check only starts if `remaining_a` and `remaining_b` are the same size.
		return true
	end

	--[==[
	Recursively compare two values that may be tables, and returns true if all key-value pairs are structurally equivalent. Note that this handles arbitrary nesting of subtables (including recursive nesting) to any depth, for keys as well as values.

	If `include_mt` is true, then metatables are also compared. If `raw` is true, then metamethods are not used during the comparison.]==]
	function export.deepEquals(a, b, include_mt, raw)
		return is_equivalent(a, b, {}, include_mt, raw and raw_pairs or pairs)
	end
	deep_equals = export.deepEquals
end

do
	local function get_nested(t, k, ...)
		if t == nil then
			return nil
		elseif select("#", ...) ~= 0 then
			return get_nested(t[k], ...)
		end
		return t[k]
	end

	--[==[
	Given a table and an arbitrary number of keys, will successively access subtables using each key in turn, returning the value at the final key. For example, if {t} is { {[1] = {[2] = {[3] = "foo"}}}}, {export.getNested(t, 1, 2, 3)} will return {"foo"}.

	If no subtable exists for a given key value, returns nil, but will throw an error if a non-table is found at an intermediary key.]==]
	function export.getNested(t, ...)
		if t == nil or select("#", ...) == 0 then
			error("Must provide a table and at least one key.")
		end
		return get_nested(t, ...)
	end
end

do
	local function set_nested(t, v, k, ...)
		if select("#", ...) == 0 then
			t[k] = v
			return
		end
		local next_t = t[k]
		if next_t == nil then
			-- If there's no next table while setting nil, there's nothing more to do.
			if v == nil then
				return
			end
			next_t = {}
			t[k] = next_t
		end
		return set_nested(next_t, v, ...)
	end

	--[==[
	Given a table, value and an arbitrary number of keys, will successively access subtables using each key in turn, and sets the value at the final key. For example, if {t} is { {} }, {export.setNested(t, "foo", 1, 2, 3)} will modify {t} to { {[1] = {[2] = {[3] = "foo"} } } }.

	If no subtable exists for a given key value, one will be created, but the function will throw an error if a non-table value is found at an intermediary key.

	Note: the parameter order (table, value, keys) differs from functions like rawset, because the number of keys can be arbitrary. This is to avoid situations where an additional argument must be appended to arbitrary lists of variables, which can be awkward and error-prone: for example, when handling variable arguments ({{lua|...}}) or function return values.]==]
	function export.setNested(t, ...)
		if t == nil or select("#", ...) < 2 then
			error("Must provide a table and at least one key.")
		end
		return set_nested(t, ...)
	end
end

--[==[
Given a list and a value to be found, return true if the value is in the array portion of the list. Comparison is by value, using `deepEquals`.]==]
function export.contains(list, x, options)
	if options and options.key then
		x = options.key(x)
	end
	local i = 0
	while true do
		i = i + 1
		local v = list[i]
		if v == nil then
			return false
		elseif options and options.key then
			v = options.key(v)
		end
		if deep_equals(v, x) then
			return true
		end
	end
end
contains = export.contains

--[==[
Given a general table and a value to be found, return true if the value is in
either the array or hashmap portion of the table. Comparison is by value, using
`deepEquals`.]==]
function export.tableContains(t, x)
	for _, v in pairs(t) do
		if deep_equals(v, x) then
			return true
		end
	end
	return false
end

--[==[
Given a `list` and a `new_item` to be inserted, append the value to the end of the list if not already present
(or insert at an arbitrary position, if `options.pos` is given; see below). Comparison is by value, using {deepEquals}.

`options` is an optional table of additional options to control the behavior of the operation. The following options are
recognized:
* `pos`: Position at which insertion happens (i.e. before the existing item at position `pos`).
* `key`: Function of one argument to return a comparison key, as with {deepEquals}. The key function is applied to both
		 `item` and the existing item in `list` to compare against, and the comparison is done against the results.
		 This is useful when inserting a complex structure into an existing list while avoiding duplicates.
* `combine`: Function of three arguments (the existing item, the new item and the position, respectively) to combine an
			 existing item with `new_item`, when `new_item` is found in `list`. If unspecified, the existing item is
			 left alone.

Return {false} if entry already found, {true} if inserted.

For compatibility, `pos` can be specified directly as the third argument in place of `options`, but this is not
recommended for new code.

NOTE: This function is O(N) in the size of the existing list. If you use this function in a loop to insert several
items, you will get O(M*(M+N)) behavior, effectively O((M+N)^2). Thus it is not recommended to use this unless you are
sure the total number of items will be small. (An alternative for large lists is to insert all the items without
checking for duplicates, and use {removeDuplicates()} at the end.)]==]
function export.insertIfNot(list, new_item, options)
	if type(options) == "number" then
		options = {pos = options}
	end
	if options and options.combine then
		local new_key
		-- Don't use options.key and options.key(new_item) or new_item in case the key is legitimately false or nil.
		if options.key then
			new_key = options.key(new_item)
		else
			new_key = new_item
		end
		local i = 0
		while true do
			i = i + 1
			local item, key = list[i]
			if item == nil then
				break
			elseif options.key then
				key = options.key(item)
			else
				key = item
			end
			if deep_equals(key, new_key) then
				local retval = options.combine(item, new_item, i)
				if retval ~= nil then
					list[i] = retval
				end
				return false
			end
		end
	elseif contains(list, new_item, options) then
		return false
	end
	local pos = options and options.pos
	if pos then
		insert(list, pos, new_item)
	else
		insert(list, new_item)
	end
end
insert_if_not = export.insertIfNot

--[==[
Finds key for specified value in a given table. Roughly equivalent to reversing the key-value pairs in the table:
* {reversed_table = { [value1] = key1, [value2] = key2, ... }}
and then returning {reversed_table[value]}. Comparison is by value, using `deepEquals`.

Only reliable if there is just one key with the specified value. Otherwise, the function returns the first key found,
and the output is unpredictable.]==]
function export.keyFor(t, x)
	for k, v in pairs(t) do
		if deep_equals(v, x) then
			return k
		end
	end
	return nil
end

do
	local types
	local function get_types()
		types = invert{
			"number",
			"boolean",
			"string",
			"table",
			"function",
			"thread",
			"userdata"
		}
		get_types = nil
		return types
	end
	
	local function compare(key1, key2)
		return key1 < key2
	end
	
	-- The default sorting function used in export.keysToList if `keySort` is not given.
	local function default_compare(key1, key2)
		local type1, type2 = type(key1), type(key2)
		if type1 ~= type2 then
			-- If the types are different, sort numbers first, functions last, and all other types alphabetically.
			return (types or get_types())[type1] < types[type2]
		-- `string_sort` fixes a bug in < which causes all codepoints above U+FFFF to be treated as equal.
		elseif type1 == "string" then
			return string_sort(key1, key2)
		elseif type1 == "number" then
			return key1 < key2
		-- Attempt to compare tables, in case there's a metamethod.
		elseif type1 == "table" then
			local success, result = pcall(compare, key1, key2)
			if success then
				return result
			end
		-- Sort true before false.
		elseif type1 == "boolean" then
			return key1
		end
		return false
	end

	--[==[
	Return a list of the keys in a table, sorted using either the default table.sort function or a custom keySort function.

	If there are only numerical keys, `export.numKeys` is probably faster.]==]
	function export.keysToList(t, keySort)
		local list, i = {}, 0
		for key in pairs(t) do
			i = i + 1
			list[i] = key
		end
		-- Use specified sort function, or otherwise `default_compare`.
		sort(list, keySort or default_compare)
		return list
	end
	keys_to_list = export.keysToList
end

--[==[
Iterates through a table, with the keys sorted using the keysToList function.

If there are only numerical keys, `export.sparseIpairs` is probably faster.]==]
function export.sortedPairs(t, keySort)
	local list, i = keys_to_list(t, keySort), 0
	return function()
		i = i + 1
		local key = list[i]
		if key ~= nil then
			return key, t[key]
		end
	end
end

do
	-- Loader.
	local function really_reverse_ipairs(...)
		really_reverse_ipairs = require(function_module).reverseIter(ipairs)
		return really_reverse_ipairs(...)
	end
	
	local function reverse_ipairs_func(t, i)
		if i > 1 then
			i = i - 1
			-- Use rawget and check `v` is not nil, in case `t` has been modified during the loop.
			local v = rawget(t, i)
			if v ~= nil then
				return i, v
			end
		end
	end
	
	function export.reverseIpairs(t)
		-- If there's no metatable, it's faster and cheaper to iterate backwards over the array part via direct table access.
		-- Note: not safe to use #t for the initial value of `i`, as it can be unpredictable if there's a hash part.
		if getmetatable(t) == nil then
			return reverse_ipairs_func, t, table_len(t, true) + 1
		end
		-- Otherwise, there could be an __ipairs metamethod (which may be hidden if the real metatable uses __metatable), so we have to assume there is one.
		-- These can return arbitrary values (i.e. the first value might not be `i`), so use reverseIter in [[Module:fun]] to guarantee accuracy.
		return really_reverse_ipairs(t)
	end
end

local function getIteratorValues(i, j , step, t_len)
	i = (i and i < 0 and t_len - i + 1) or i or (step and step < 0 and t_len) or 1
	j = (j and j < 0 and t_len - j + 1) or j or (step and step < 0 and 1) or t_len
	step = step or (j < i and -1) or 1
	if (
		i == 0 or i % 1 ~= 0 or
		j == 0 or j % 1 ~= 0 or
		step == 0 or step % 1 ~= 0
	) then
		error("Arguments i, j and step must be non-zero integers.")
	end
	return i, j, step
end

--[==[
Given an array `list` and function `func`, iterate through the array applying {func(r, k, v)}, and returning the result,
where `r` is the value calculated so far, `k` is an index, and `v` is the value at index `k`. For example,
{reduce(array, function(a, b) return a + b end)} will return the sum of `array`.

Optional arguments:
* `i`: start index; negative values count from the end of the array
* `j`: end index; negative values count from the end of the array
* `step`: step increment
These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or
backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:
# No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
# step=-1 results in backward iteration from the end to the start in steps of 1.
# i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
# j=-3 results in forward iteration from the start to the 3rd last index.
# j=-3, step=-1 results in backward iteration from the end to the 3rd last index.
Note: directionality generally only matters for `reduce`, but values of step > 1 (or step < -1) still affect the return value
of `apply`.]==]
function export.reduce(t, func, i, j, step)
	local t_len = table_len(t)
	i, j, step = getIteratorValues(i, j , step, t_len)
	local ret = t[i]
	for k = i + step, j, step do
		ret = func(ret, k, t[k])
	end
	return ret
end

--[==[
Given an array `list` and function `func`, iterate through the array applying {func(k, v)} (where `k` is an index, and
`v` is the value at index `k`), and return an array of the resulting values. For example,
{apply(array, function(a) return 2*a end)} will return an array where each member of `array` has been doubled.

Optional arguments:
* `i`: start index; negative values count from the end of the array
* `j`: end index; negative values count from the end of the array
* `step`: step increment
These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or
backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:
# No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
# step=-1 results in backward iteration from the end to the start in steps of 1.
# i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
# j=-3 results in forward iteration from the start to the 3rd last index.
# j=-3, step=-1 results in backward iteration from the end to the 3rd last index.
Note: directionality makes the most difference for `reduce`, but values of step > 1 (or step < -1) still affect the return
value of `apply`.]==]
function export.apply(t, func, i, j, step)
	local t_new = deep_copy(t)
	local t_new_len = table_len(t_new)
	i, j, step = getIteratorValues(i, j , step, t_new_len)
	for k = i, j, step do
		t_new[k] = func(k, t_new[k])
	end
	return t_new
end

--[==[
Given an array `list` and function `func`, iterate through the array applying {func(k, v)} (where `k` is an index, and
`v` is the value at index `k`), and returning whether the function is true for all iterations.

Optional arguments:
* `i`: start index; negative values count from the end of the array
* `j`: end index; negative values count from the end of the array
* `step`: step increment
These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or
backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:
# No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
# step=-1 results in backward iteration from the end to the start in steps of 1.
# i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
# j=-3 results in forward iteration from the start to the 3rd last index.
# j=-3, step=-1 results in backward iteration from the end to the 3rd last index.]==]
function export.all(t, func, i, j, step)
	local t_len = table_len(t)
	i, j, step = getIteratorValues(i, j , step, t_len)
	for k = i, j, step do
		if not func(k, t[k]) then
			return false
		end
	end
	return true
end

--[==[
Given an array `list` and function `func`, iterate through the array applying {func(k, v)} (where `k` is an index, and
`v` is the value at index `k`), and returning whether the function is true for at least one iteration.

Optional arguments:
* `i`: start index; negative values count from the end of the array
* `j`: end index; negative values count from the end of the array
* `step`: step increment
These must be non-zero integers. The function will determine where to iterate from, whether to iterate forwards or
backwards and by how much, based on these inputs (see examples below for default behaviours).

Examples:
# No values for i, j or step results in forward iteration from the start to the end in steps of 1 (the default).
# step=-1 results in backward iteration from the end to the start in steps of 1.
# i=7, j=3 results in backward iteration from indices 7 to 3 in steps of 1 (i.e. step=-1).
# j=-3 results in forward iteration from the start to the 3rd last index.
# j=-3, step=-1 results in backward iteration from the end to the 3rd last index.]==]
function export.any(t, func, i, j, step)
	local t_len = table_len(t)
	i, j, step = getIteratorValues(i, j , step, t_len)
	for k = i, j, step do
		if not not (func(k, t[k])) then
			return true
		end
	end
	return false
end

--[==[
Joins an array with serial comma and serial conjunction, normally {"and"}. An improvement on {mw.text.listToText},
which doesn't properly handle serial commas.

Options:
* `conj`: Conjunction to use; defaults to {"and"}.
* `punc`: Punctuation to use; default to {","}.
* `dontTag`: Don't tag the serial comma and serial {"and"}. For error messages, in which HTML cannot be used.]==]
function export.serialCommaJoin(seq, options)
	local length = table_len(seq)
	if length == 0 then
		return ""
	elseif length == 1 then
		return seq[1]
	end
	
	if length == 2 then
		return seq[1] .. " " .. (options and options.conj or "and") .. " " .. seq[2]
	end

	local conj, punc, dont_tag
	if options then
		conj = options.conj or "and"
		punc = options.punc or ","
		dont_tag = options.dontTag
	else
		conj, punc = "and", ","
	end
	
	local comma
	if dont_tag then
		comma = punc
		conj = " " .. conj .. " "
	else
		comma = "<span class=\"serial-comma\">" .. punc .. "</span>"
		conj = "<span class=\"serial-and\"> " .. conj .. "</span> "
	end
	
	return concat(seq, punc .. " ", 1, length - 1) .. comma .. conj .. seq[length]
end

--[==[
A function which works like `table.concat`, but respects any `__index` metamethod. This is useful for data loaded via `mw.loadData`.]==]
function export.concat(t, sep, i, j)
	local list, k = {}, 0
	while true do
		k = k + 1
		local v = t[k]
		if v == nil then
			return concat(list, sep, i, j)
		end
		list[k] = v
	end
end

--[==[
Concatenate all values in the table that are indexed by a number, in order.
* {sparseConcat{ a, nil, c, d }}  =>  {"acd"}
* {sparseConcat{ nil, b, c, d }}  =>  {"bcd"}]==]
function export.sparseConcat(t, sep, i, j)
	local list, k = {}, 0
	for _, v in sparse_ipairs(t) do
		k = k + 1
		list[k] = v
	end
	return concat(list, sep, i, j)
end

--[==[
Values of numeric keys in array portion of table are reversed: { { "a", "b", "c" }} -> { { "c", "b", "a" }}]==]
function export.reverse(t)
	local list, t_len = {}, table_len(t)
	for i = t_len, 1, -1 do
		list[t_len - i + 1] = t[i]
	end
	return list
end
table_reverse = export.reverse

function export.reverseConcat(t, sep, i, j)
	return concat(table_reverse(t), sep, i, j)
end

--[==[
Invert a list. For example, {invert({ "a", "b", "c" })} -> { { a = 1, b = 2, c = 3 }}]==]
function export.invert(list)
	local map, i = {}, 0
	while true do
		i = i + 1
		local v = list[i]
		if v == nil then
			return map
		end
		map[v] = i
	end
end
invert = export.invert

--[==[
Convert `list` (a table with a list of values) into a set (a table where those values are keys instead). This is a useful
way to create a fast lookup table, since looking up a table key is much, much faster than iterating over the whole list
to see if it contains a given value.

By default, each item is given the value true. If the optional parameter `value` is a function or functor, then it is called
as an iterator, with the list index as the first argument, the item as the second (which will be used as the key), plus any
additional arguments passed to {listToSet}; the returned value is used as the value for that list item. If `value` is anything
else, then it is used as the fixed value for every item.]==]
function export.listToSet(list, value, ...)
	local set, i, callable = {}, 0
	if value == nil then
		value = true
	else
		callable = is_callable(value)
	end
	while true do
		i = i + 1
		local item = list[i]
		if item == nil then
			return set
		end
		if callable then
			set[item] = value(i, item, ...)
		else
			set[item] = value
		end
	end
end

--[==[
Return true if all keys in the table are consecutive integers starting at 1.]==]
function export.isArray(t)
	local i = 0
	for _ in pairs(t) do
		i = i + 1
		if t[i] == nil then
			return false
		end
	end
	return true
end

--[==[
Add a list of aliases for a given key to a table. The aliases must be given as a table.]==]
function export.alias(t, k, aliases)
	for _, alias in pairs(aliases) do
		t[alias] = t[k]
	end
end

return export