My main project at the moment is a custom MediaWiki installation that hosts a hyper-textual guide to Star Trek. It is something similar to the widely-known Memory Alpha, surely in a smaller size, but it is in italian language and based on a standard modern version of MediaWiki. Furthermore, this project is heavily integrated with WikiBase, a substantial extension to MediaWiki that provides database-like capabilities and it is the cornerstone of open data for Wikipedia being the base of WikiData.
MediaWiki and LUA
Basic usage of data from WikiBase is very easy, but any complex access to data needs custom coding in LUA. LUA is provided as a language for customising MediaWiki through an extension called Scribunto. WikiBase comes with a LUA Scribunto interface that allows easy manipulation.
There is a manual on how to use the LUA interface that lists and explain all the object model, but IMHO fails in providing enough examples.
So I started the painful experience, so familiar to all developers, of trials and errors till I found a way to make everything work.
While the LUA code I was writing was specific to my project, I started thinking it could also be made more generic and maybe useful to other users, so i thought a way to share it.
Being on a public Wiki, the code was accessible to anyone (but editable only to members, just as a passing comment) so already shared, in sense. Nevertheless, the truth is, in practical terms, that our wiki is about a niche product of a niche of the entertainment industry; incidentally the wiki is also written in italian, reducing its potential readership even more.
I needed a better idea.
How to share
WikiTrek is a GitHub "organization" already, so the first thing I thought was to publish the LUA code on a dedicated repository. At first I started to copy and paste the code manually, so each time I made a modification to LUA code, I copied and pasted it.
It was very inefficient and prone to errors.
A better approach would have been to transfer data automatically from the wiki pages to GitHub.
First important thing was not to reinvent the wheel, so I started looking for a ready-to-use PHP library to interact with GitHub and found PHP GitHub API pretty soon. It is an excellent library to perform any kind of operations on GitHub using ready-made PHP classes.
Finding the right way
So I started playing around with PHP GitHub API and, in a couple of hours (documentation and examples of that library might benefit from some work, but this is another story), I had a proof-of-concept small application to manipulate GitHub via that API.
My original plan was to write a MW Maintenance script and run it on the server periodically using cron or maybe MW own scheduler.
The more I looked into this solution, though, the more complicated it became.
After a few days of alternatively banging my head against the problem and thinking about it, I came up with a different solution.
I started looking into extension development and found out there is a class of extensions using Hooks allowing to inject custom PHP code at certain event. Most prominent example of these events is PageContentSaveComplete and that was all I needed: by intercepting a page when it is saved, I could have taken the content to send it to GitHub.
Bingo.
Building the extension
First step was to familiarise myself with the extension development: MediaWiki documentation on the topic is vast and – as unfortunately often happens with that project – not completely structured.
I found the two most important thing to master:
- the intricacy of Extension registration using the
extension.jsonfile and - the handling of custom settings via configuration variables
Once I understood the basics of extension development, I had to tackle some issues about the on the working of the extension:
- I wanted only some pages to be transferred, so I found a way to select the pages belonging to
Module:namespace only - All the pages with code on them ends up in the
Module:namespace, so I needed to select only specific pages, otherwise I would have copied code belonging to other extensions or template or whatever, most likely code non even written by contributors to the site. I resolved to do so by introducing a custom keyword to look for in the page. - I implemented a mechanism to not save the pages if the edit is marked as "Minor"
Firstly I hard-coded all of the above for testing purpose: by the end of the projects, I made all options user-customisable variables that can be handled in Localsettings.php
Closing words
The result of all this effort is an extension that I called PageToGitHub (or P2G, in brief). I don't expect this to be used by someone else as it address a very specific feature that is hardly to be found elsewhere.
Nevertheless I am very satisfied with it because, apart of solving the original issue I had, the journey to create it helped me to improve my (otherwise very limited) PHP skills and guided me to better understand inner MediaWiki mechanisms.
Beside, of course, allowing me to write this post and start my humble contribution to DEV community.
Top comments (1)
It's always wonderful to see LUA in all caps!