Wikipedia:Guide to Scribbling

"Shh! I'm reading about how to Scribble templates."

This is the Guide toScribbling.Scribbling, also known as Luafication, is the act of writing a template, or converting a template, so that it usesthe Scribunto extensiontoMediaWiki.The Scribunto extension[a]was developed byTim StarlingandVictor Vasiliev,and allows for embedding scripting languages in MediaWiki. Currently the only supported scripting language isLua.This Guide aims to give you a broad overview of Scribbling, and pointers to further information in various places.

Scribbled templates come in two parts: the template itself and one or more back-endmodules— in theModule:namespace — that contain programs that are run onthe wiki serversto generate the wikitext that the template expands to. The template invokes a function within a module using a newparser functionnamed{{#invoke:}}.

The idea of Scribbling is to improve template processing performance. Scribbling eliminates any need fortemplate parser function programmingusing parser functions such as{{#if}},{{#ifeq}},{{#switch}}and{{#expr}}.All of this is instead done in the module, in a language that was actually designed to be a programming language, rather than a template system onto which was bolted various extensions over time to try to make it into a programming language.[b]Scribbling also eliminates any need for templates to expand to further templates and potentially hit theexpansion depth limit.A fully Scribbled template should never need totranscludeother templates.[c]

The language in which modules are written is Lua. Unlike the template parser function system, Lua was actually designed not only to be a proper programming language, but also to be a programming language that is suitable for what is known asembedded scripting.Modules in MediaWiki are an example of embedded scripts. There are several embedded scripting languages that could have been used, includingREXXandtcl;and indeed the original aim of Scribunto was to make available a choice of such languages. At the moment, however, only Lua is available.

The official reference manual for Lua isIerusalimschy, de Figueiredo & Celes 2006.It's a reference, not a tutorial. Consult it if you want to know the syntax or semantics for something. For a tutorial, see eitherIerusalimschy 2006(Ierusalimschy 2003is also available, although it is of course out of date.) orJung & Brown 2007.The downsides to these books are that quite a lot of the things that they tell you about have no bearing upon using Lua in MediaWiki modules. You don't need to know how to install Lua and how to integrate its interpreter into a program or run it standalone. The MediaWiki developers have done all of that. Similarly, a lot of the Lua library functions are, for security, not available in modules. (For example, it's not possible to do file I/O or to make operating system calls in MediaWiki modules.) So, much of what these books explain about Lua standard library functions and variables that come with the language is either irrelevant or untrue here.

The original API specification — the Lua standard library functions and variables that are supposed to be available in modules — is given atMW:Extension:Scribunto/API specification.However, even that is untrue. What you'llactuallyhave available is documented inMW:Extension:Scribunto/Lua reference manual,which is a cut down version of the 1st Edition Lua manual that has been edited down and modified by Tim Starling to bring it more into line with the reality of Scribbling. Again, though, this is a reference manual, not a tutorial.

The things in Lua that you will mostly be concerned with, writing Scribbled templates, aretables,stringsnumbers,booleans,nil,if...then...else...end,while...do...end,for...in...do...end(generatedfor),for...do...end(numericalfor),repeat...until,function...end,local,return,break,expressions and the various operators (including#,..,the arithmetic operators+,-,*,/,^,and%), and thestring,math,andmwglobal tables (i.e. libraries).

Template structure

edit

This is simple. Your template comprises one expansion of{{#invoke:}}in the usual case. Here is{{Harvard citation}},for example:

<includeonly>{{#invoke:Footnotes|harvard_citation
|bracket_left=(
|bracket_right=)
}}</includeonly><noinclude>
{{documentation}}
<!-- Add categories to the /doc subpage, interwikis to Wikidata, not here -->
</noinclude>

If you find yourself wanting to use other templates within your template, or to use template parser functions, or indeed anything at all other than{{#invoke:}}and possibly somevariablesas its arguments,then you are using the wrong approach.

Module basics

edit

Overall structure

edit

Let's consider a hypothetical module,Module:Population.(SeeModule:Population clocksfor a similar, but more complex, module.) It can be structured in one of two ways:

A named local table

edit
localp={}

functionp.India(frame)
return"1,21,01,93,422 people at (nominally) 2011-03-01 00:00:00 +0530"
end

returnp

An unnamed table generated on the fly

edit
return{
India=function(frame)
return"1,21,01,93,422 people at (nominally) 2011-03-01 00:00:00 +0530"
end
}

Execution

edit

The execution of a module by{{#invoke:}}is actually twofold:

  1. The module is loaded and the entire script is run. This loads up any additional modules that the module needs (using therequire()function), builds the (invocable) functions that the module will provide to templates, and returns a table of them.
  2. The function named in{{#invoke:}}is picked out of the table built in phase 1 and called, with the arguments supplied to the templateandthe arguments supplied to{{#invoke:}}(more on whichlater).

The first Lua script does phase 1 fairly explicitly. It creates alocal variablenamedpon line 1, initialized to a table; builds and adds a function to it (lines 3–5), by giving the function the nameIndiain the table named byp(functionp.Indiabeing the same as sayingp["India"]=function[d]); and then returns (on line 7) the table as the last line of the script. To expand such a script with more (invocable) functions, one adds them between thelocalstatement at the top and thereturnstatement at the bottom. (Non-invocablelocalfunctions can be addedbeforethelocalstatement.) The local variable doesn't have to be namedp.It could be named any valid Lua variable name that you like.pis simply conventional for this purpose, and is also the name that you can use to test the script in the debug console of the Module editor.

The second Lua script does the same thing, but more "idiomatically". Instead of creating a named variable as a table, it creates an anonymous table on the fly, in the middle of thereturnstatement, which is the only (executed during the first phase) statement in the script. TheIndia=function(frame)...endon lines 2–4 creates an (also anonymous) function and inserts it into the table under the nameIndia.To expand such a script with more (invocable) functions, one adds them as further fields in the table. (Non-invocablelocalfunctions can, again, be addedbeforethereturnstatement.)

In both cases, the template code that one writes is{{#invoke:Population|India}}toinvokethe function namedIndiafrom the moduleModule:Population.Also note thatfunctionbuildsa function, as an object, to be called. It doesn'tdeclareit, as you might be used to from other programming languages, and the function isn't executed until itiscalled.

One can do more complex things than this, of course. For example: One can declare other local variables in addition top,to hold tables of data (such as lists of Language or country names), that the module uses. But this is thebasicstructure of a module. You make a table full of stuff, and return it.

Receiving template arguments

edit

An ordinary function in Lua can take an (effectively) arbitrary number of arguments. Witness this function fromModule:Wikitextthat can be called with anywhere between zero and three arguments:

functionz.oxfordlist(args,separator,ampersand)

Functions called by{{#invoke:}}are special. They expect to be passed exactly one argument, a table that is called aframe(and so is conventionally given the parameter nameframein the parameter list of the function). It's called aframebecause, unfortunately, the developers chose to name it for their convenience. It's named after an internal structure within the code of MediaWiki itself, which itsort ofrepresents.[e]

This frame has a (sub-)table within it, namedargs.It also has a means for accessing itsparent frame(again, named after a thing in MediaWiki). The parent framealsohas a (sub-)table within it, also namedargs.

  • The arguments in the (child, one supposes) frame — i.e. the value of theframeparameter to the function — are the arguments passed to{{#invoke:}}within the wikitext of your template.So, for example, if you were to write{{#invoke:Population|India|a|b|class="popdata"}}in your template then the arguments sub-table of the child frame would be (as written in Lua form){"a","b",class="popdata"}.
  • The arguments in the parent frame are the argumentspassed to your template when it was transcluded.So, for example, were the user of your template to write{{Population of India|c|d|language=Hindi}}then the arguments sub-table of the parent frame would be (as written in Lua form){"c","d",language="Hindi"}.

A handy programmers' idiom that you can use, to make this all a bit easier, is to have local variables named (say)configandargsin your function, that point to these two argument tables. See this, fromModule:WikidataCheck:

functionp.wikidatacheck(frame)
localpframe=frame:getParent()
localconfig=frame.args-- the arguments passed BY the template, in the wikitext of the template itself
localargs=pframe.args-- the arguments passed TO the template, in the wikitext that transcludes the template

Everything inconfigis thus an argument thatyouhave specified, in your template, that you can reference with code such asconfig[1]andconfig["class"].These will be things that tell your module function its "configuration" (e.g. a CSS class name that can vary according to what template is used).

Everything inargsis thus an argument thatthe user of the templatehas specified, where it was transcluded, that you can reference with code such asargs[1]andargs["language"].These will be the normal template arguments, as documented on your template's/docpage.

See{{other places}}and{{other ships}}for two templates that both do{{#invoke:Other uses|otherX|x}}but do so with different arguments in place of thex,thereby obtaining different results from one single common Lua function.

For both sets of arguments, the name and value of the argument are exactly as in the wikitext, except that leading and trailing whitespace in named parameters is discounted. This has an effect on your code if you decide to support or employ transclusion/invocation argument names that aren't valid Lua variable names. You cannot use the "dot" form of table lookup in such cases. For instance:args.author-firstis, as you can see from the syntax colourization here, not a reference to an|author-first=argument, but a reference to an|author=argument and afirstvariable with the subtraction operator in the middle. To access such an argument, use the "square bracket" form of table lookup:args["author-first"].

Named arguments are indexed in theargstable by their name strings, of course. Positional arguments (whether as the result of an explicit1=or otherwise) are indexed in theargstables by number, not by string.args[1]is not the same asargs["1"],and the latter is effectively unsettable from wikitext.

Finally, note that Lua modules can differentiate between arguments that have been used in the wikitext and simply set to an empty string, and arguments that aren't in the wikitext at all. The latter don't exist in theargstable, and any attempt to index them will evaluate tonil.Whereas the formerdoexist in the table and evaluate to an empty string,"".

Errors

edit

Let's get one thing out of the way right at the start:Script erroris a hyperlink. You can put the mouse pointer on it and click.

We've become so conditioned by our (non-Scribbled) templates putting out error messages in red that we think that the Scribunto "Script error" error message is nothing but more of the same. It isn't. If you haveJavaScriptenabled in your WWW browser, it will pop up a window giving the details of the error, a call backtrace, and even hyperlinks that will take you to the location of the code where the error happened in the relevant module.

You can cause an error to happen by calling theerror()function.

Tips and tricks

edit

Arguments tables are "special".

edit

For reasons that are out of the scope of this Guide,[f]theargssub-table of a frame is not quite like an ordinary table. It starts out empty, and it is populated with arguments as and when you execute code that looks for them.[g](It's possible to make tables that work like this in a Lua program, using things calledmetatables.That, too, is outwith the scope of this Guide.)

An unfortunate side-effect of this is that some of the normal Lua table operators don't work on anargstable. The length operator,#,will not work, and neither will the functions in Lua'stablelibrary. These only work with standard tables, and fail when presented with the specialargstable. However, thepairs()andipairs()functions will both work, as code to make their use possible has been added by the developers.

Copy table contents into local variables.

edit

A name in Lua is either an access of a local variable or a table lookup.[3]math.flooris a table lookup (of the string"floor") in the (global)mathtable, for example. Table lookups are slower, at runtime, than local variable lookups. Table lookups in tables such as theargstable withits "specialness"are alotslower.

A function in Lua can have up to 250 local variables.[4]So make liberal use of them:

  • If you callmath.floormany times, copy it into a local variable and use that instead:[4]
    localfloor=math.floor
    locala=floor((14-date.mon)/12)
    localy=date.year+4800-a
    localm=date.mon+12*a-3
    returndate.day+floor((153*m+2)/5)+365*y+floor(y/4)-floor(y/100)+floor(y/400)-2432046
    
  • Don't useargs.somethingover and over. Copy it into a local variable and use that:
    localTab=args.tab
    
    (Even theargsvariable itself is a way to avoid looking up"args"in theframetable over and over.)

When copying arguments into local variables there are two useful things that you can do along the way:

  • Thealternative names for the same argumenttrick. If a template argument can go by different names — such as uppercase and lowercase forms, or different English spellings — then you can use Lua'soroperator to pick the highest priority name that is actually supplied:
    localTitle=args.titleorargs.encyclopaediaorargs.encyclopediaorargs.dictionary
    localISBN=args.isbn13orargs.isbnorargs.ISBN
    

This works for two reasons:

    • nilis the same asfalseas far asoris concerned.
    • Lua'soroperator has what are known as "shortcut" semantics. If the left-hand operand evaluates to something that isn'tfalseornil,it doesn't bother even working out the value of the right-hand operand. (So whilst that first example may at first glance look like it does four lookups, in the commonest case, where|title=is used with the template, it in fact only actually does one.)
  • Thedefault to empty stringtrick. Sometimes the fact that an omitted template argument isnilis useful. Other times, however, it isn't, and you want the behaviour of missing arguments being empty strings. A simpleor""at the end of an expression suffices:
    localID=args.idorargs.IDorargs[1]or""
    

Don't expand templates, even though you can.

edit

If local variables are cheap and table lookups are expensive, then template expansion is way above your price bracket.

Avoidframe:preprocess()like the plague. Nested template expansion using MediaWiki's preprocessor is what we're trying to get away from, after all. Most things that you'd do with that are done more simply, more quickly, and more maintainably, with simple Lua functions.

Similarly, avoid things like usingw:Template:ISO 639 name aze(deleted August 2020) to store what is effectively an entry in a database. Reading it would be a nested parser call with concomitant database queries, all to map a string onto another string. Put a simple straightforward data table in your module, like the ones inModule:Wikt-lang.

Notes

edit
  1. ^The name "Scribunto" is Latin. "scribunto"is third person pluralfutureactiveimperativeof "scribere"and means" they shall write "."scribble"is of course an English word derived from that Latin word, viaMediaeval Latin"scribillare".[1]
  2. ^For an idea of what "bolted-on" connotes when it comes to software design, see theFlintstonescartoons where the rack of ribs from the Drive-Thru is so heavy that it causes the Flintstones' car to fall on its side.
  3. ^It may need, until such time as the whole of the specified API for Scribunto is available to modules, to transcludemagic words.Seethe tips and tricks section.Magic words are not templates, however.
  4. ^The inventors of the language call thissyntactic sugar.[2]
  5. ^In MediaWiki proper, there are more than two frames.
  6. ^If you want to know, go and read about how MediaWiki, in part due to the burden laid upon it by the old templates-conditionally-transcluding-templates system, doeslazy evaluationof template arguments.
  7. ^Don't be surprised, therefore, if you find a call backtrace showing a call to some other module in what you thought was an ordinary template argument reference. That will be because expansion of that argument involved expanding another Scribbled template.

References

edit

Cross-references

edit

Citations

edit
  • "scribble".Merriam-Webster's Collegiate Dictionary: Eleventh Edition.Merriam-Webster's Collegiate Dictionary(11th ed.). Merriam-Webster. 2003. p. 1116.ISBN9780877798095.
  • Ierusalimschy, Roberto; de Figueiredo, Luiz Henrique; Celes, Waldemar (12 May 2011)."Passing a Language through the Eye of a Needle".Queue.9(5). Association for Computing Machinery. ACM 1542-7730/11/0500.
  • Ierusalimschy, Roberto (December 2008)."Lua Performance Tips"(PDF).In de Figueiredo, Luiz Henrique; Celes, Waldemar; Ierusalimschy, Roberto (eds.).Lua Programming Gems.Lua.org.ISBN978-85-903798-4-3.

Further reading

edit