Vincent Massol wrote:
On Sep 30, 2008, at 9:22 AM, Anca Paula Luca wrote:
Vincent Massol wrote:
Very good! :)
Some small comments:
* Not sure why you call the class HelloWorldContextualizer instead of HelloWorldVelocityContextInitializer Just for a shorter name and because that component does not actually initialize velocity context, it just uses that to get access to the VelocityContext. So its semantic is not that of initializing velo context but of contextualizing the HelloWorld. Happily enough, this hijack method won't last. If you think we really must, I can change it.
The main reason for my comment is that the notion of Context and Contextualizer exists in component based programming and in plexus and it's something different. For example we could have added a Contextualizer interface with a single method:
contextualize(ExecutionContext context);
It's not very important but my fear is that readers will use your example when they write code and would lead to badly named classes.
They most probably will. Will change.
BTW I see you've used that in Watch components so that should be modified too IMO.
* Might be good to have a small section about the fact that components remove the need for static code (for the best!) since you can have a component with a "singleton" lifecycle. Maybe we should have a better section about the <instantiation- strategy> and what can be done with it, because eliminating the static code is not the components' main benefit and I'd like to keep the tutorial to "what you really really need to know" stage, with links for further info.
Thanks -Vincent
The rest should be for Sergiu's expertise since he's the one that contributed those last three paragraphs.
* "and since we don't want to add the whole core and all its dependencies as a dependency of a simple lightweight component, ". Actually the real problem is that this would create a circular dependency since the core uses the components. * "In the component world, the current request information is held in an execution context. This is actually more powerful than the old XWiki context, as it is a generic execution context, and you can create one anytime you want and use it anyway you want. And you don't have to manually pass it around with all method calls, as execution contexts are managed by the Execution component, which you can use just like any other XWiki component.". We need to specify that it's a ThreadLocal implementation (meaning you get a new instance of the Execution context per thread). This means also that it's initialized by default at all XWiki entry points (i.e. where threads are created: Servlet, Portlet, Lucene thread, Stats thread, etc). * "If you need some functionality from the old core, consider rewriting that part as a new component first, and then use that new component from your code.". I'd rather ask to discuss it on the list so that we find a solution collaboratively
I even learnt something! Good idea with the old xwikicontext cast to a Map, hasn't thought of that :)
Excellent work Anca.
Big +1 to link it from our dev practices document.
Thanks -Vincent
On Sep 29, 2008, at 8:33 PM, Anca Paula Luca wrote:
Hi devs,
while developing my XWiki watch component, I've written a quick start tutorial on XWiki components, which I would like to add in the dedicated section on platform.xwiki.org.
It is currently in draft version at http://dev.xwiki.org/xwiki/bin/view/Drafts/CreatingComponents . Please give it a critical look and signal any errors that might have slipped or send improvement suggestions.
Here's my +1 for taking it out of draft version, WDYT?
Happy coding, Anca Luca
devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs