Difference between revisions of "Module:Arguments"

From Amanita Research
Jump to navigation Jump to search
(memoize nils, add type checks for frame and options, use string library functions instead of mw.text.trim and mw.ustring.find, define four different tidyVal functions to avoid checking options every time)
m (tweak comment)
Line 55: Line 55:
  
 
-- Get the arguments from the frame object if available. If the frame object is not available, we are being called
 
-- 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 put the args in a special table so we can differentiate them.
+
-- 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
 
local fargs, pargs, luaArgs
 
if type(frame.args) == 'table' and type(frame.getParent) == 'function' then
 
if type(frame.args) == 'table' and type(frame.getParent) == 'function' then

Revision as of 03:13, 12 December 2013

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.
  • Arguments are fetched as needed, which can help avoid (some) problems with <ref>...</ref> tags.
  • 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 Template: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 in the code {{#invoke:Side box|main|text=Some text}} 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 in the reference list at the bottom of the article. If your module proceeds to omit the reference tag from the output, you will end up with a phantom reference – a reference that appears in the reference list but without any 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 = {}

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

-- 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 up the args and metaArgs tables. args will be the one accessed from functions, and metaArgs will hold the actual arguments.
	-- The metatable connects the two together.
	local args, metaArgs, metatable = {}, {}, {}
	setmetatable(args, metatable)

	-- 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

	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 precendence.
		-- 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
				local metaArgsVal = metaArgs[key]
				if metaArgsVal == nil or metaArgsVal == nilArg then
					local tidiedVal = tidyVal(key, val)
					if tidiedVal == nil then
						metaArgs[key] = nilArg
					else
						metaArgs[key] = tidiedVal
					end
				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, and are only fetched from the
	-- argument tables once. Nil arguments are also memoized using the nilArg variable in order to increase
	-- performance. 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 fargs and pargs 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)
		local val = metaArgs[key]
		if val ~= nil then
			if val == nilArg then
				return nil
			else
				return val
			end
		end
		for _, argTable in ipairs(argTables) do
			local argTableVal = tidyVal(key, argTable[key])
			if argTableVal == nil then
				metaArgs[key] = nilArg
			else
				metaArgs[key] = argTableVal
				return argTableVal
			end
		end
		return nil
	end

	metatable.__newindex = function (t, key, val)
		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
			metaArgs[key] = nilArg -- Memoize nils.
		else
			metaArgs[key] = val
		end
	end

	metatable.__pairs = function ()
		if not metatable.donePairs then
			mergeArgs(pairs, argTables)
			metatable.donePairs = true
			metatable.doneIpairs = true
		end
		return pairs(metaArgs)
	end

	metatable.__ipairs = function ()
		if not metatable.doneIpairs then
			mergeArgs(ipairs, argTables)
			metatable.doneIpairs = true
		end
		return ipairs(metaArgs)
	end

	return args
end

return arguments