Tips and Tricks
Feature-based self-hosting The current method of self-hosting in Eclipse is plug-in-based. PDE launches a second run-time workbench instance by passing an array of plug-ins that it should load. A regular Eclipse product is feature-based: during startup, it checks all the features that should be active, computes plug-ins that belong to those features, and passes the result for loading. This difference in behavior makes it complicated to self-host in scenarios where a full startup that involves features is required. PDE now supports this scenario if care is taken with the setup:
- The workspace needs to be <work-area>/plugins.
- Features must be imported into the workspace using the new 'Feature Import' wizard (they will be created in <work-area>/features).
- All plug-ins must be in the workspace (either in source or imported as binary projects WITHOUT linking).
- When launching, Run-time Workbench launcher must be configured to use features (in Plug-ins and Fragments tab).
If all these conditions are met, the runtime Eclipse instance will be launched in a way that is the closest possible approximation of a normal Eclipse startup. This facilitates testing About dialogs and other aspects that may depend on the set of installed features.
To clean or not to clean When you create a new runtime workbench launch configuration, PDE presets the Program Arguments on the launch configuration to include a -clean argument. This -clean argument clears all runtime-cached data in your runtime workbench from one invocation to the next to ensure that all the changes made in your host workbench, e.g. new Java packages were added to a plug-in project, etc., are picked up when you launch a runtime workbench.
This clearing of the cache may hinder performance if your target platform contains a large number of plug-ins.
Therefore, if you're in a situation where your target platform has a large number of plug-ins and you're at a stage where you are not actively adding/removing packages from your plug-in projects, you could remove the -clean argument from the launch configuration to improve startup time.
Importing with linking Importing external plug-ins and fragments can be time consuming and may result in large workspaces, depending on the content of the plug-ins being imported. Therefore, the 'Import External Plug-ins and Fragments' wizard gives you the option to import with linking. This means that the import operation will not copy the resources being imported into your workspace. It will simply create links to the files being imported. You will be able to browse these linked resources, as if they had been copied into your workspace. However, they are physically not there on your file system, so you will not be able to modify them. Beware of operations that depend on files being physically in your workspace, as they will not work on linked resources. Templates For a quick start, PDE provides several template plug-ins that will generate a plug-in with one or more fully-working extensions. In addition, if at any point, you would like to add a new extension from the template list (without having to generate a plug-in), you could access these extension templates directly from the manifest editor. From the 'Extensions' page of the editor, click 'Add...'. In the wizard that comes up, select Extension Templates in the left pane and choose the template of choice in the right pane. Plug-in dependency extent If you have ever looked at the list of plug-ins that your plug-in depends on and wondered why your plug-in needs a particular plug-in X, now you can easily find out why. The Compute Dependency Extent operation found on the context menu in several contexts (including manifest file Dependencies page and Dependencies view) performs a combined Java and plug-in search to find all Java types and extension points provided by plug-in X which are referenced by your plug-in. The results will be displayed in the Search view. When a type is selected in the Search results view, the References in MyPlugIn action in the context menu searches for the places in the plug-in where the selected type is referenced.
If the search returns 0 results, you should definitely remove plug-in X from your list of dependencies, as it is not being used at all, and it would just slow class loading.
The Compute Dependency Extent is also useful to check if you are using internal (non-API) classes from plug-in X, which might not be desirable.
Finding unused dependencies Minimizing a plug-in's number of dependencies is certain to improve performance. As your plug-in evolves, its list of dependencies might become stale, as it might be still containing references to plug-ins that it no longer needs. A quick way to check that all dependencies listed by your plug-in are actually used by the plug-in is to run the 'Find Unused Dependencies' utility, which is available through the context menu of the 'Dependencies' page of PDE's manifest editor. Extending the Java search scope Java Search is limited to projects in your workspace and external jars that these projects reference. If you would like to add more libraries from external plug-ins into the search: open the Plug-ins View, select a plug-in and choose Add to Java Search from the context menu. This is handy for remaining aware of other plug-ins that depend on ones you're working on.