Module:Arguments: Difference between revisions

From Zoophilia Wiki
Jump to navigationJump to search
meta>Mr. Stradivarius
fix undefined next() behaviour bug by checking for metatable.donePairs in the __index metamethod; also, format the module so it fits into 80 characters
meta>Mr. Stradivarius
memoize nils using a separate nilArgs table; this fixes a bug where nil values were being iterated over with pairs()
Line 7: Line 7:


local arguments = {}
local arguments = {}
local nilArg = {} -- Used for memoizing nil arguments in metaArgs.


-- Generate four different tidyVal functions, so that we don't have to check the
-- Generate four different tidyVal functions, so that we don't have to check the
Line 76: Line 74:
luaArgs = frame
luaArgs = frame
end
end
 
-- Set up the args and metaArgs tables. args will be the one accessed from
-- Set the order of precedence of the argument tables. If the variables are
-- functions, and metaArgs will hold the actual arguments. The metatable
-- nil, nothing will be added to the table, which is how we avoid clashes
-- connects the two together.
-- between the frame/parent args and the Lua args.
local args, metaArgs, metatable = {}, {}, {}
local argTables = {fargs}
setmetatable(args, metatable)
argTables[#argTables + 1] = pargs
argTables[#argTables + 1] = luaArgs


--[[
--[[
Line 113: Line 112:
end
end
end
end
--[[
-- Set up the args, metaArgs and nilArgs tables. args will be the one
-- accessed from functions, and metaArgs will hold the actual arguments. Nil
-- arguments are memoized in nilArgs, and the metatable connects all of them
-- together.
--]]
local args, metaArgs, nilArgs, metatable = {}, {}, {}, {}
setmetatable(args, metatable)


local function mergeArgs(iterator, tables)
local function mergeArgs(iterator, tables)
Line 124: Line 132:
for _, t in ipairs(tables) do
for _, t in ipairs(tables) do
for key, val in iterator(t) do
for key, val in iterator(t) do
local metaArgsVal = metaArgs[key]
if metaArgs[key] == nil then
if metaArgsVal == nil or metaArgsVal == nilArg then
local tidiedVal = tidyVal(key, val)
local tidiedVal = tidyVal(key, val)
if tidiedVal == nil then
if tidiedVal == nil then
metaArgs[key] = nilArg
nilArgs[key] = true
else
else
metaArgs[key] = tidiedVal
metaArgs[key] = tidiedVal
Line 136: Line 143:
end
end
end
end
-- Set the order of precedence of the argument tables. If the variables are
-- nil, nothing will be added to the table, which is how we avoid clashes
-- between the frame/parent args and the Lua args.
local argTables = {fargs}
argTables[#argTables + 1] = pargs
argTables[#argTables + 1] = luaArgs


--[[
--[[
-- Define metatable behaviour. Arguments are memoized in the metaArgs table,
-- Define metatable behaviour. Arguments are memoized in the metaArgs table,
-- and are only fetched from the argument tables once. Nil arguments are
-- and are only fetched from the argument tables once. Fetching arguments
-- also memoized using the nilArg variable in order to increase performance.
-- from the argument tables is the most resource-intensive step in this
-- Also, we keep a record in the metatable of when pairs and ipairs have
-- module, so we try and avoid it where possible. For this reason, nil
-- been called, so we do not run pairs and ipairs on fargs and pargs more
-- arguments are also memoized, in the nilArgs table. Also, we keep a record
-- than once. We also do not run ipairs on fargs and pargs if pairs has
-- in the metatable of when pairs and ipairs have been called, so we do not
-- already been run, as all the arguments will already have been copied
-- run pairs and ipairs on the argument tables more than once. We also do
-- over.
-- not run ipairs on fargs and pargs if pairs has already been run, as all
-- the arguments will already have been copied over.
--]]
--]]


metatable.__index = function (t, key)
metatable.__index = function (t, key)
--[[
-- Fetches an argument when the args table is indexed. First we check
-- to see if the value is memoized, and if not we try and fetch it from
-- the argument tables. When we check memoization, we need to check
-- metaArgs before nilArgs, as both can be non-nil at the same time.
-- If the argument is not present in metaArgs, we also check whether
-- pairs has been run yet. If pairs has already been run, we return nil.
-- This is because all the arguments will have already been copied into
-- metaArgs by the mergeArgs function, meaning that any other arguments
-- must be nil.
--]]
local val = metaArgs[key]
local val = metaArgs[key]
if metatable.donePairs or val ~= nil then
if val ~= nil then
--[[
return val
-- We have either memoized the argument already, or pairs has been
elseif metatable.donePairs or nilArgs[key] then
-- called, meaning that mergeArgs has already copied all of the
return nil
-- available arguments into the metaArgs table. We need to check for
-- pairs as we can't memoize nils to the metaArgs table while pairs
-- is iterating. Adding new instances of nilArg to the metaArgs
-- table while pairs is iterating over it produces undefined
-- behaviour in the next() function.
--]]
if val == nilArg then
return nil
else
return val
end
end
end
for _, argTable in ipairs(argTables) do
for _, argTable in ipairs(argTables) do
local argTableVal = tidyVal(key, argTable[key])
local argTableVal = tidyVal(key, argTable[key])
if argTableVal == nil then
if argTableVal == nil then
metaArgs[key] = nilArg
nilArgs[key] = true
else
else
metaArgs[key] = argTableVal
metaArgs[key] = argTableVal
Line 186: Line 187:


metatable.__newindex = function (t, key, val)
metatable.__newindex = function (t, key, val)
-- This function is called when a module tries to add a new value to the
-- args table, or tries to change an existing value.
if options.readOnly then
if options.readOnly then
error(
error(
Line 201: Line 204:
)
)
elseif val == nil then
elseif val == nil then
metaArgs[key] = nilArg -- Memoize nils.
--[[
-- If the argument is to be overwritten with nil, we need to erase
-- the value in metaArgs, so that __index, __pairs and __ipairs do
-- not use a previous existing value, if present; and we also need
-- to memoize the nil in nilArgs, so that the value isn't looked
-- up in the argument tables if it is accessed again.
--]]
metaArgs[key] = nil
nilArgs[key] = true -- Memoize nils.
else
else
metaArgs[key] = val
metaArgs[key] = val
Line 208: Line 219:


metatable.__pairs = function ()
metatable.__pairs = function ()
-- Called when pairs is run on the args table.
if not metatable.donePairs then
if not metatable.donePairs then
mergeArgs(pairs, argTables)
mergeArgs(pairs, argTables)
Line 213: Line 225:
metatable.doneIpairs = true
metatable.doneIpairs = true
end
end
return function (t, k)
return pairs(metaArgs)
local nk, val = next(metaArgs, k)
if val == nilArg then
val = nil
end
return nk, val
end
end
end


metatable.__ipairs = function ()
metatable.__ipairs = function ()
-- Called when ipairs is run on the args table.
if not metatable.doneIpairs then
if not metatable.doneIpairs then
mergeArgs(ipairs, argTables)
mergeArgs(ipairs, argTables)
metatable.doneIpairs = true
metatable.doneIpairs = true
end
end
return function (t, i)
return ipairs(metaArgs)
local val = metaArgs[i + 1]
if val == nil then
return nil
elseif val == nilArg then
val = nil
end
return i + 1, val
end, nil, 0
end
end



Revision as of 10:30, 18 April 2014

This module provides easy processing of arguments passed from #invoke. It is a meta-module, meant for use by other modules, and should not be called from #invoke directly. Its features include:

  • Easy trimming of arguments and removal of blank arguments.
  • Arguments can be passed by both the current frame and by the parent frame at the same time. (More details below.)
  • Arguments can be passed in directly from another Lua module or from the debug console.
  • Most features can be customized.

Basic use

First, you need to load the module. It contains one function, named getArgs.

local getArgs = require('Module:Arguments').getArgs

In the most basic scenario, you can use getArgs inside your main function. The variable args is a table containing the arguments from #invoke. (See below for details.)

local getArgs = require('Module:Arguments').getArgs
local p = {}

function p.main(frame)
	local args = getArgs(frame)
	-- Main module code goes here.
end

return p

However, the recommended practice is to use a function just for processing arguments from #invoke. This means that if someone calls your module from another Lua module, you don't have to have a frame object available, which improves performance.

local getArgs = require('Module:Arguments').getArgs
local p = {}

function p.main(frame)
	local args = getArgs(frame)
	return p._main(args)
end

function p._main(args)
	-- Main module code goes here.
end

return p

If you want multiple functions to use the arguments, and you also want them to be accessible from #invoke, you can use a wrapper function.

local getArgs = require('Module:Arguments').getArgs

local p = {}

local function makeInvokeFunc(funcName)
	return function (frame)
		local args = getArgs(frame)
		return p[funcName](args)
	end
end

p.func1 = makeInvokeFunc('_func1')

function p._func1(args)
	-- Code for the first function goes here.
end

p.func2 = makeInvokeFunc('_func2')

function p._func2(args)
	-- Code for the second function goes here.
end

return p

Options

The following options are available. They are explained in the sections below.

local args = getArgs(frame, {
	trim = false,
	removeBlanks = false,
	valueFunc = function (key, value)
		-- Code for processing one argument
	end,
	frameOnly = true,
	parentOnly = true,
	parentFirst = true,
	wrappers = {
		'Template:A wrapper template',
		'Template:Another wrapper template'
	},
	readOnly = true,
	noOverwrite = true
})

Trimming and removing blanks

Blank arguments often trip up coders new to converting MediaWiki templates to Lua. In template syntax, blank strings and strings consisting only of whitespace are considered false. However, in Lua, blank strings and strings consisting of whitespace are considered true. This means that if you don't pay attention to such arguments when you write your Lua modules, you might treat something as true that should actually be treated as false. To avoid this, by default this module removes all blank arguments.

Similarly, whitespace can cause problems when dealing with positional arguments. Although whitespace is trimmed for named arguments coming from #invoke, it is preserved for positional arguments. Most of the time this additional whitespace is not desired, so this module trims it off by default.

However, sometimes you want to use blank arguments as input, and sometimes you want to keep additional whitespace. This can be necessary to convert some templates exactly as they were written. If you want to do this, you can set the trim and removeBlanks arguments to false.

local args = getArgs(frame, {
	trim = false,
	removeBlanks = false
})

Custom formatting of arguments

Sometimes you want to remove some blank arguments but not others, or perhaps you might want to put all of the positional arguments in lower case. To do things like this you can use the valueFunc option. The input to this option must be a function that takes two parameters, key and value, and returns a single value. This value is what you will get when you access the field key in the args table.

Example 1: This function preserves whitespace for the first positional argument, but trims all other arguments and removes all other blank arguments.

local args = getArgs(frame, {
	valueFunc = function (key, value)
		if key == 1 then
			return value
		elseif value then
			value = mw.text.trim(value)
			if value ~= '' then
				return value
			end
		end
		return nil
	end
})

Example 2: This function removes blank arguments and converts all arguments to lower case, but doesn't trim whitespace from positional parameters.

local args = getArgs(frame, {
	valueFunc = function (key, value)
		if not value then
			return nil
		end
		value = mw.ustring.lower(value)
		if mw.ustring.find(value, '%S') then
			return value
		end
		return nil
	end
})

Note: The above functions will fail if passed input that is not of type string or nil. This might be the case if you use the getArgs function in the main function of your module, and that function is called by another Lua module. In this case, you will need to check the type of your input. This is not a problem if you are using a function specially for arguments from #invoke (i.e. you have p.main and p._main functions, or something similar).

Examples 1 and 2, with type checking
Example 1:
local args = getArgs(frame, {
	valueFunc = function (key, value)
		if key == 1 then
			return value
		elseif type(value) == 'string' then
			value = mw.text.trim(value)
			if value ~= '' then
				return value
			else
				return nil
			end
		else
			return value
		end
	end
})

Example 2:

local args = getArgs(frame, {
	valueFunc = function (key, value)
		if type(value) == 'string' then
			value = mw.ustring.lower(value)
			if mw.ustring.find(value, '%S') then
				return value
			else
				return nil
			end
		else
			return value
		end
	end
})

Also, please note that the valueFunc function is called more or less every time an argument is requested from the args table, so if you care about performance you should make sure you aren't doing anything inefficient with your code.

Frames and parent frames

Arguments in the args table can be passed from the current frame or from its parent frame at the same time. To understand what this means, it is easiest to give an example. Let's say that we have a module called Module:ExampleArgs. This module prints the first two positional arguments that it is passed.

Module:ExampleArgs code
local getArgs = require('Module:Arguments').getArgs
local p = {}

function p.main(frame)
	local args = getArgs(frame)
	return p._main(args)
end

function p._main(args)
	local first = args[1] or ''
	local second = args[2] or ''
	return first .. ' ' .. second
end

return p

Module:ExampleArgs is then called by Template:ExampleArgs, which contains the code {{#invoke:ExampleArgs|main|firstInvokeArg}}. This produces the result firstInvokeArg.

Now if we were to call Template:ExampleArgs, the following would happen:

Code Result
{{ExampleArgs}} firstInvokeArg
{{ExampleArgs|firstTemplateArg}} firstInvokeArg
{{ExampleArgs|firstTemplateArg|secondTemplateArg}} firstInvokeArg secondTemplateArg

There are three options you can set to change this behaviour: frameOnly, parentOnly and parentFirst. If you set frameOnly, then only arguments passed from the current frame will be accepted. If you set parentOnly, then only arguments passed from the parent frame will be accepted. And if you set parentFirst then arguments will be passed from both the current and parent frames, but the parent frame will have priority over the current frame. Here are the results in terms of Template:ExampleArgs:

frameOnly
Code Result
{{ExampleArgs}} firstInvokeArg
{{ExampleArgs|firstTemplateArg}} firstInvokeArg
{{ExampleArgs|firstTemplateArg|secondTemplateArg}} firstInvokeArg
parentOnly
Code Result
{{ExampleArgs}}
{{ExampleArgs|firstTemplateArg}} firstTemplateArg
{{ExampleArgs|firstTemplateArg|secondTemplateArg}} firstTemplateArg secondTemplateArg
parentFirst
Code Result
{{ExampleArgs}} firstInvokeArg
{{ExampleArgs|firstTemplateArg}} firstTemplateArg
{{ExampleArgs|firstTemplateArg|secondTemplateArg}} firstTemplateArg secondTemplateArg

Notes:

  1. If you set both the frameOnly and parentOnly options, the module won't fetch any arguments at all from #invoke. This is probably not what you want.
  2. In some situations, a parent frame may not be available, e.g. if getArgs is passed the parent frame rather than the current frame. In this case, only the frame arguments will be used (unless parentOnly is set, in which case no arguments will be used) and the parentFirst and frameOnly options will have no effect.

Wrappers

The wrappers option is used to specify a limited number of templates as wrapper templates, that is, templates whose only purpose is to call a module. If the module detects that it is being called from a wrapper template, it will only check for arguments in the parent frame; otherwise it will only check for arguments in the frame passed to getArgs. This allows modules to be called by either #invoke or through a wrapper template without the loss of performance associated with having to check both the frame and the parent frame for each argument lookup.

For example, the only content of {{Side box}} (excluding content in <noinclude>...</noinclude> tags) is {{#invoke:Side box|main}}. There is no point in checking the arguments passed directly to the #invoke statement for this template, as no arguments will ever be specified there. We can avoid checking arguments passed to #invoke by using the parentOnly option, but if we do this, then #invoke will not work from other pages either. If this were the case, the |text=Some text parameter in the code {{#invoke:Side box|main|text=<var>Some text</var>}} would be ignored completely, no matter what page it was used from. By using the wrappers option to specify 'Template:Side box' as a wrapper, we can make {{#invoke:Side box|main|text=Some text}} work from most pages, while still not requiring that the module check for arguments on the Template:Side box page itself.

Wrappers can be specified either as a string, or as an array of strings.

local args = getArgs(frame, {
	wrappers = 'Template:Wrapper template'
})
local args = getArgs(frame, {
	wrappers = {
		'Template:Wrapper 1',
		'Template:Wrapper 2',
		-- Any number of wrapper templates can be added here.
	}
})

Notes:

  1. The module will automatically detect if it is being called from a wrapper template's /sandbox subpage, so there is no need to specify sandbox pages explicitly.
  2. The wrappers option effectively changes the default of the frameOnly and parentOnly options. If, for example, parentOnly were explicitly set to false with wrappers set, calls via wrapper templates would result in both frame and parent arguments being loaded, though calls not via wrapper templates would result in only frame arguments being loaded.
  3. If the wrappers option is set and no parent frame is available, the module will always get the arguments from the frame passed to getArgs.

Writing to the args table

Sometimes it can be useful to write new values to the args table. This is possible with the default settings of this module. (However, bear in mind that it is usually better coding style to create a new table with your new values and copy arguments from the args table as needed.)

args.foo = 'some value'

It is possible to alter this behaviour with the readOnly and noOverwrite options. If readOnly is set, then it is not possible to write any values to the args table at all. If noOverwrite is set, then it is possible to add new values to the table, but it is not possible to add a value if it would overwrite any arguments that are passed from #invoke.

Ref tags

This module uses metatables to fetch arguments from #invoke. This allows access to both the frame arguments and the parent frame arguments without using the pairs() function. This can help if your module might be passed <ref>...</ref> tags as input.

As soon as <ref>...</ref> tags are accessed from Lua, they are processed by the Mediawiki software and the reference will appear on the page at the same position as the <references /> tag in the page's source wikitext. If your module proceeds to omit the <ref> tag from the output, you will end up with a phantom reference: a reference that appears in the references list but without any corresponding superscript citation number linking to it. This has been a problem with modules that use pairs() to detect whether to use the arguments from the frame or the parent frame, as those modules automatically process every available argument.

This module solves this problem by allowing access to both frame and parent frame arguments, while still only fetching those arguments when it is necessary. The problem will still occur if you use pairs(args) elsewhere in your module, however.

Known limitations

The use of metatables also has its downsides. Most of the normal Lua table tools won't work properly on the args table, including the # operator, the next() function, and the functions in the table library. If using these is important for your module, you should use your own argument processing function instead of this module.


-- This module provides easy processing of arguments passed to Scribunto from
-- #invoke. It is intended for use by other Lua modules, and should not be
-- called from #invoke directly.

local libraryUtil = require('libraryUtil')
local checkType = libraryUtil.checkType

local arguments = {}

-- Generate four different tidyVal functions, so that we don't have to check the
-- options every time we call it.

local function tidyValDefault(key, val)
	if type(val) == 'string' then
		val = val:match('^%s*(.-)%s*$')
		if val == '' then
			return nil
		else
			return val
		end
	else
		return val
	end
end

local function tidyValTrimOnly(key, val)
	if type(val) == 'string' then
		return val:match('^%s*(.-)%s*$')
	else
		return val
	end
end

local function tidyValRemoveBlanksOnly(key, val)
	if type(val) == 'string' then
		if val:find('%S') then
			return val
		else
			return nil
		end
	else
		return val
	end
end

local function tidyValNoChange(key, val)
	return val
end

function arguments.getArgs(frame, options)
	checkType('getArgs', 1, frame, 'table', true)
	checkType('getArgs', 2, options, 'table', true)
	frame = frame or {}
	options = options or {}

	--[[
	-- Get the arguments from the frame object if available. If the frame object
	-- is not available, we are being called from another Lua module or from the
	-- debug console, so assign the args to a new variable so we can
	-- differentiate them.
	--]]
	local fargs, pargs, luaArgs
	if type(frame.args) == 'table' and type(frame.getParent) == 'function' then
		if not options.parentOnly then
			fargs = frame.args
		end
		if not options.frameOnly then
			pargs = frame:getParent().args
		end
		if options.parentFirst then
			fargs, pargs = pargs, fargs
		end
	else
		luaArgs = frame
	end
	
	-- Set the order of precedence of the argument tables. If the variables are
	-- nil, nothing will be added to the table, which is how we avoid clashes
	-- between the frame/parent args and the Lua args.	
	local argTables = {fargs}
	argTables[#argTables + 1] = pargs
	argTables[#argTables + 1] = luaArgs

	--[[
	-- Generate the tidyVal function. If it has been specified by the user, we
	-- use that; if not, we choose one of four functions depending on the
	-- options chosen. This is so that we don't have to call the options table
	-- every time the function is called.
	--]]
	local tidyVal = options.valueFunc
	if tidyVal then
		if type(tidyVal) ~= 'function' then
			error(
				"bad value assigned to option 'valueFunc'"
					.. '(function expected, got '
					.. type(tidyVal)
					.. ')',
				2
			)
		end
	elseif options.trim ~= false then
		if options.removeBlanks ~= false then
			tidyVal = tidyValDefault
		else
			tidyVal = tidyValTrimOnly
		end
	else
		if options.removeBlanks ~= false then
			tidyVal = tidyValRemoveBlanksOnly
		else
			tidyVal = tidyValNoChange
		end
	end

	--[[
	-- Set up the args, metaArgs and nilArgs tables. args will be the one
	-- accessed from functions, and metaArgs will hold the actual arguments. Nil
	-- arguments are memoized in nilArgs, and the metatable connects all of them
	-- together.
	--]]
	local args, metaArgs, nilArgs, metatable = {}, {}, {}, {}
	setmetatable(args, metatable)

	local function mergeArgs(iterator, tables)
		--[[
		-- Accepts multiple tables as input and merges their keys and values
		-- into one table using the specified iterator. If a value is already
		-- present it is not overwritten; tables listed earlier have precedence.
		-- We are also memoizing nil values, but those values can be
		-- overwritten.
		--]]
		for _, t in ipairs(tables) do
			for key, val in iterator(t) do
				if metaArgs[key] == nil then
					local tidiedVal = tidyVal(key, val)
					if tidiedVal == nil then
						nilArgs[key] = true
					else
						metaArgs[key] = tidiedVal
					end
				end
			end
		end
	end

	--[[
	-- Define metatable behaviour. Arguments are memoized in the metaArgs table,
	-- and are only fetched from the argument tables once. Fetching arguments
	-- from the argument tables is the most resource-intensive step in this
	-- module, so we try and avoid it where possible. For this reason, nil
	-- arguments are also memoized, in the nilArgs table. Also, we keep a record
	-- in the metatable of when pairs and ipairs have been called, so we do not
	-- run pairs and ipairs on the argument tables more than once. We also do
	-- not run ipairs on fargs and pargs if pairs has already been run, as all
	-- the arguments will already have been copied over.
	--]]

	metatable.__index = function (t, key)
		--[[
		-- Fetches an argument when the args table is indexed. First we check
		-- to see if the value is memoized, and if not we try and fetch it from
		-- the argument tables. When we check memoization, we need to check
		-- metaArgs before nilArgs, as both can be non-nil at the same time.
		-- If the argument is not present in metaArgs, we also check whether
		-- pairs has been run yet. If pairs has already been run, we return nil.
		-- This is because all the arguments will have already been copied into
		-- metaArgs by the mergeArgs function, meaning that any other arguments
		-- must be nil.
		--]]
		local val = metaArgs[key]
		if val ~= nil then
			return val
		elseif metatable.donePairs or nilArgs[key] then
			return nil
		end
		for _, argTable in ipairs(argTables) do
			local argTableVal = tidyVal(key, argTable[key])
			if argTableVal == nil then
				nilArgs[key] = true
			else
				metaArgs[key] = argTableVal
				return argTableVal
			end
		end
		return nil
	end

	metatable.__newindex = function (t, key, val)
		-- This function is called when a module tries to add a new value to the
		-- args table, or tries to change an existing value.
		if options.readOnly then
			error(
				'could not write to argument table key "'
					.. tostring(key)
					.. '"; the table is read-only',
				2
			)
		elseif options.noOverwrite and args[key] ~= nil then
			error(
				'could not write to argument table key "'
					.. tostring(key)
					.. '"; overwriting existing arguments is not permitted',
				2
			)
		elseif val == nil then
			--[[
			-- If the argument is to be overwritten with nil, we need to erase
			-- the value in metaArgs, so that __index, __pairs and __ipairs do
			-- not use a previous existing value, if present; and we also need
			-- to memoize the nil in nilArgs, so that the value isn't looked
			-- up in the argument tables if it is accessed again.
			--]]
			metaArgs[key] = nil
			nilArgs[key] = true -- Memoize nils.
		else
			metaArgs[key] = val
		end
	end

	metatable.__pairs = function ()
		-- Called when pairs is run on the args table.
		if not metatable.donePairs then
			mergeArgs(pairs, argTables)
			metatable.donePairs = true
			metatable.doneIpairs = true
		end
		return pairs(metaArgs)
	end

	metatable.__ipairs = function ()
		-- Called when ipairs is run on the args table.
		if not metatable.doneIpairs then
			mergeArgs(ipairs, argTables)
			metatable.doneIpairs = true
		end
		return ipairs(metaArgs)
	end

	return args
end

return arguments