Difference between revisions of "Platform UI/Plan/4.4"
(→Exposing Eclipse4 in the Workbench)
(→Logging and Tracing)
|Line 65:||Line 65:|
=== Logging and Tracing ===
=== Logging and Tracing ===
We need to decide on our logging and tracing strategies, and which APIs to use (ex, just use Equinox Logging and DebugTrace?). Then we need to apply that strategy to many of our new Eclipse4 plugins.
We need to decide on our logging and tracing strategies, and which APIs to use (ex, just use Equinox Logging and DebugTrace ?). Then we need to apply that strategy to many of our new Eclipse4 plugins.
=== Scheduling Work ===
=== Scheduling Work ===
Revision as of 10:28, 27 June 2013
- 1 Planning page for Luna
- 2 Themes
- 2.1 Exposing Eclipse4 in the Workbench
- 2.2 IEclipseContext and DI Use
- 2.3 Fragments and Extension Points
- 2.4 Graduate Relevant Eclipse4 tools
- 2.5 Eclipse Application Services
- 2.6 Workbench Services Decoupling
- 2.7 Application Startup and Lifecycle
- 2.8 CSS Enhancements
- 2.9 Stability and Performance
Planning page for Luna
Draft to develop a plan for our work in Luna
The general themes that we should focus on in Luna.
Exposing Eclipse4 in the Workbench
Migration path for 3.x plugins
- Figure out how a 3.x extension point can take advantage of contexts and DI
Hosting Eclipse4 Models
The Workbench (a.k.a compatibility layer) manipulates the model to host 3.x views and editors on it. In Luna we'd like to see Eclipse4 parts and other contributions contributed to the Workbench and provide 3.x wrappers to correctly interact with them in the Workbench.
- bug 376486 - Eclipse 4 IDE not extendable via fragments or processors
Menus, Commands, and Handlers
Handlers in the Workbench are currently clustered on their own IEclipseContext just off of the application context. The handlers in Eclipse4 can be instantiated at any MContext level, and that controls their activation and enablement much finer grained. The goal is to expose this distribution on handlers contributed through the Workbench.
IEclipseContext and DI Use
The IEclipseContext represent a runtime hierarchy of information. The goal is to have a regular pattern and predictable use, so that looking up information at any point in the hierarchy will return the appropriate answer.
Some mechanisms it supports we need to rationalize:
- setting the value for a variable, and then use context.getActiveLeaf().get(variable)
- using context.modify(variable, value) and then using context.get(variable)
- setting a variable in the application context that points to a ContextFunction. context.get(variable) will apply the ContextFunction to the current hierarchy.
- Review what data is in the context and whether it belongs elsewhere such as transient data.
- Investigate how to provide bulk updates to contexts efficiently
- Investigate threading problems
Fragments and Extension Points
The fragment processing needs some more design work. The goal is to simplify the creation of fragments and to apply them in a standard way that is predictable.
- bug 278631 - Extensions and E4
Some enhancements in that area:
- bug 324954 - [Model] Allow model fragments to use XPath-Syntax to decide where to contribute
- bug 396974 - Multiple Parents for MenuContribution item
It would be nice if there was a tool where you could create your structure, select elements, and generate a fragment.
Graduate Relevant Eclipse4 tools
There are tools still in the E4 incubator that we should look at graduating or moth balling in Luna. Some currently under consideration:
The Lightweight CSS editor will probably need some work before it can graduate, as it would bring in the XText runtime. See bug 410841 - Explorer editing CSS with orion
Eclipse Application Services
See E4/Eclipse Application Services as a starting point, although our understanding has changed some of the categories.
Logging and Tracing
We need to decide on our logging and tracing strategies, and which APIs to use (ex, just use Equinox Logging and DebugTrace or SLF4J?). Then we need to apply that strategy to many of our new Eclipse4 plugins.
There was some discussion about using equinox concurrent support, and re-framing our scheduling/progress in terms of that. We need to analyze the difference between Jobs and IWorkbenchRunnableWithProgress and the concurrent support, and then look at our Eclipse4 level support. i.e. Do we use Jobs + provide Progress? Does Jobs need to be in terms of equinox concurrent? Do we need a parallel progress pattern with equinox concurrent?
We need to provide the fully functioning Eclipse4 save lifecycle, that the Workbench save lifecycle can be implemented on top of.
Workbench Services Decoupling
It's always been our goal to break up many of the Workbench services so they can be consumed by Eclipse4 applications without consuming the entire Workbench. Some examples:
- ISharedImages service
- Progress view service
- Error log service
- Preference management
- Properties management
- New wizards
Our starting priority here was the Error log and Progress view.
- bug 404727 - Make 3.x workbench resources available for e4 apps
- bug 401655 - Create e4 based Progress View
Application Startup and Lifecycle
The goal is that the Eclipse4 application startup and lifecycle make sense, and the Workbench startup fits in the appropriate place.
For example, it's reasonable that the Workbench startup should create a useful model that can then be passed to the Eclipse4 application startup. There are a number of lifecycle hooks that should probably be made available in the Eclipse4 startup.
There are also more general discussions about lifecycle:
- bug 398837 - We need to revamp the e4 life cycle handling
- bug 392903 - Easier and more powerful lifecycle handling for Windows, Perspectives
The goal here is to improve our CSS story.
- The workflow of how we load and use CSS within eclipse
- Enhancements to use it more throughout SWT.
- How the CSS interacts with the Model
- Moving the CSS engine up to the latest Batik.
- Provide some reasonable bridge interactions between CSS and our older Colors and Fonts properties.
Stability and Performance
We still have some immediate issues to address in terms of stability and performance. See our Platform UI 4.3.1 bug list.
We will continue to focus on regressions, and improvements in the UX workflows. Some bugs we've already collected for 4.4.
We have existing performance bugs:
- bug 385394 - Performance issue regarding enabled state handling for menuContributions
- bug 373294 - Improve startup time
- bug 406042 - Performance degrade on createEditorActionBars() call
In addition, we are going to do a performance pass early this year, probably in the M3 timeframe.