|Help developers to navigate|
|Table of Contents|
This page should help developers navigate who want to support Semantic MediaWiki development or develop extensions to SMW can use a number of resources to get started. Of course, it is strongly recommended to be familiar with the usage of SMW as well.
Moreover, there is an SMW architecture guide that provides a basic introduction to the main ideas and concepts in SMW from a developer perspective, which should be useful for reading and writing SMW-related code.
This document should help newcomers and developers to navigate around Semantic MediaWiki and its development environment.
The main objective of the
Semantic MediaWiki software is to provide "semantic" functions on top of MediaWiki to enable machine-reading of wiki-content and allow structured content to be queried and displayed by means of employing different backends including:
SQLStoreto be used as default storage and query engine for small and mid-size wikis
ElasticStorerecommended to large wiki farms which need to scale or for users with a requirement to combine structured and unstructured searches
SPARQLStorefor advanced users that have an extended requirement to work with a triple store and linked data
Getting started ...
- Read this document
- Install git
@dev(with development happening against the master branch)
composer testlocally (see the test section) and verify that your installation and test environment are setup correctly so that you would find something like "OK, but incomplete, skipped, or risky tests! Tests ..." at the end as output (after the test run)
Creating a pull request
- Send a PR with subject [first pr] to the
Semantic MediaWikirepository and verify that your git setup works and you are able to replicate changes against the master branch
- Get yourself familiar with the Travis-CI environment and observe how a PR triggers CI jobs and review the output of those jobs (important when a job doesn't pass and you need to find the cause for a failure)
Preparing a PR
- Create a PR with your changes and send it to the
- Observe whether tests are failing or not, and when there are failing identify what caused them to fail
- In case your PR went green without violating any existing tests, go back to your original PR and add tests that covers the newly introduced behaviour (see the difference for unit and integration tests)
- Rebase and repost your PR with the newly added tests and verify that they pass on all voting CI jobs
In an event that you encountered a problem, ask or create a ticket.
The general policy of the
Semantic MediaWiki software and the development thereof is:
- No MediaWiki tables are modified or altered, any data that needs to be stored persistently is relying on the Semantic MediaWiki's own
database schema(writing to the cache is an exception)
- No MediaWiki classes are modified, patched, or otherwise changed
- Only publicly available
APIinterfaces are used to extend MediaWiki with Semantic MediaWiki functions
Some conventions to help developers and the project to maintain a consistent product and helps to create testable components where classes have a smaller footprint and come with a dedicated responsibility.
- The top-level namespace is
SMWand each component should be placed in a namespace that represents the main responsibility of the component
PSR-4is used for resolving classes and namespaces in the
includesis the legacy folder that doesn't necessarily follow any of the policies or conventions mentioned in this document)
- Development happens against the master branch (see also the release process) and will be release according the the available release plan, backports should be cherry-picked and merged into the targeted branch
- Semantic MediaWiki tries to depend only on a selected pool of MediaWiki core classes (
Language... ) to minimize the potential for breakage during release changes
- It is expected that each new class and functionality is covered by corresponding unit tests and if the functionality spans into different components integration tests are required as well to ensure that the behaviour is tested across components and produces deterministic and observable outputs.
classhas a defined responsibility and boundary
- Dependency injection goes before inheritance, meaning that all objects used in a class should be injected.
- Instance creation (e.g.
new Foo( ... )) is delegated to a factory service
- Object interaction with MediaWiki objects should be done using accessors in the
- A factory service should avoid using conditionals (
if ... then ...) to create an instance
- Instance creation and dependency injection are done using a service locator or dependency builder
type hintingconsistently throughout a repository is vital to ensure class contracts can be appropriately followed
- Trying to follow
Single responsibility principleand applying
inversion of control(i.e dependency injection, factory pattern, service locator pattern) is a best practice approach
- Newly added functionality is expected to be accompanied by unit and integration test to ensure that its operation is verifiable and doesn't interfere with existing services
- Newly introduced features (or enhancements) that alter existing behaviour need to be guarded by a behaviour switch (or flag) allowing to restore any previous behaviour and need to be accompanied by integration tests
- To improve the readability of classes in terms of what is public and what are internals (not to be exposed outside of the class boundary), functions are ordered by its visibility where
protectedwhich comes before
Datamodelcontains the most essential architectural choice of Semantic MediaWiki for the management of its data including
Database schemaand table definitions in Semantic MediaWiki
Hacking by examples
- Creating annotations and storing data
- Querying data
- Writing a result printer
- Best practices for developing an extension and a list of hooks to extend the functionality of Semantic MediaWiki
- Register a custom datatype or predefined property
- Extending consistency checks on a property page
- Extending property annotators for core predefined properties, see also the Semantic Extra Special Properties extension that provides a development space for deploying other predefined (special) properties
- Extending constraints and their checks
- Working with and changing the table schema of Semantic MediaWiki
Semantic MediaWiki software alone deploys ~7000 tests (as of March 2019) which are required to pass before changes can be merged into the repository.
Tests are commonly divided into unit and integration tests where unit tests represent an isolated unit (or component) to be tested and normally doesn't require a database or repository connection. The integration test on the other hand is as test that requires other components and directly interacts with MediaWiki and its services which is why 80% of the CI time for a test run is spend on executing integration test as they will a run a full cycle (parsing, storing, reading, HTML generating etc.)
For an introduction on "How to use
PHPUnit" and "How to write integration tests using
JSONScript" this document contains relevant details.
Continues integration (CI)
The project uses Travis-CI to run its tests on different platforms with different services enabled to provide a wide range of environments including MySQL, SQLite, and Postgres.
- Contributing to Semantic MediaWiki
- Help:Writing testable code
- Help:Unit tests
- Integration tests
- MediaWiki resources for developers
Web applications with open user communities are specifically threatened by security vulnerabilities. SMW developers are responsible for taking specific care to avoid vulnerabilities of all kinds. Every developer should carefully read the MediaWiki security guidelines for developers.
For more information, please read security and software vulnerabilities.