Also see Naming conventions.
Magnolia contributors are kindly requested to follow these simple conventions below:
- Standard Java coding conventions (download the PDF, because the HTML version is very badly formatted in some places)
- Tab policy: spaces only
- Indentation: 4 spaces
- Maximum line width: 120 (no strong rule here - don't put exceptions declarations on new lines just to respect this, for instance)
- We preferably don't use abbreviations in classnames (only exception to that rule is MgnlSomething)
- Read http://dev.magnolia-cms.com/~gjoseph/on-code-readability
Also ensure that you have the right license headers in place !
See the file attached for Eclipse Preferences (which you can simply import and everything should be done).
- Code Templates
.. or see Checkstyle for Eclipse. (similar instructions apply for IntelliJ)
Preferences>Java>Code Style>Code Templates>Automatically add comments... is not used as there are different headers for the different editions. The alternative is ... (todo)
- Class-level documentation is mandatory.
- we do use @version $Id$
- we don't use @author tags any longer as we're usually not interested in the one who had create the file (SCM could tell
you if required)
- First sentence of any Javadoc comment must finish with punctuation. (This is/was needed for summary pages/tables to include the first sentence next to the described element)
- See http://dev.magnolia-cms.com/~gjoseph/on-code-readability
Style of JUnit-Tests
With Magnolia 5 we witched to JUnit 4. As it is fully backwards compatible this does not enforce you to write your tests in the new v4-style (e.g. use @Test, @Before etc.).
We set up the following rule:
- all new test will be written in JUnit 4 style
- existing tests can/should be updated if you're touching them anyway
- existing superclasses for tests (especially in core) cannot that easily be updated as there might be lots of subclasses affected by that change. We'll lift them to v4 style in a consolidated action
We don't explicitly set serialVersionUID's as we don't want to decide ourselves when these id's should be updated. We rely on the JVM doing that for us (being aware that this id's could potentially differ from JVM to JVM).