| Re: [p2-dev] IRequiredCapability internal? |
|
That would be fine with me. We have an insulation layer we can slide these in to if they were public.
Hi John and Marcus, I would like to see some API for extracting name, namespace, and range too, but not as an interface implemented by the requirement as such. The requirement is just an _expression_. Helper methods that can extract different kinds of information from an _expression_ should be kept separate. Perhaps the static helper methods on the internal RequiredCapability could be moved to a public location? MetadataFactory perhaps? Or a new RequirementHelper class? - thomas On 03/09/2010 08:09 PM, John Arthorne wrote:
However, if we know that *some* requirements can be expressed in the form namespace/name/range, then I'm not sure why we don't offer an API such as IRequiredCapability to obtain that information. That doesn't prevent us from having more general or specialized forms of requirements that are *not* represented as IRequiredCapability. If you have an IU that is known to be an OSGi bundle, then it seems reasonable to want to be able to introspect the metadata to find out things like the bundle/package imports via the p2 metadata API. I'm quite possibly missing something here though... John On Tue, Mar 9, 2010 at 12:15 PM, Thomas M Houser <tmhouser@xxxxxxxxxx> wrote: Not really. I'm charged with retargeting our code to Eclipse 3.6 (formerly targeted to 3.5 provisional API). I can only respond with generalities. We have build tooling for constructing product offerings. This build tooling generates other metadata based on p2 metadata. Therefore it reads p2 metadata and reasons about its structure. I can't give you any details, but I'll ask the engineers who use this API to respond. In general, though, I think it is kind of weird that you allow a client to construct a IRequiredCapability (via MetadataFactory.createRequiredCapability()), but do not allow a client to get the attributes of this concrete subclass.
Hi Thomas, Given that an IRequirement no longer must contain name, namespace, and range, but instead can be an arbitrary _expression_ applied on an IInstallableUnit, perhaps we can help you find an alternative solution. Can you say something about why you need access to the requirement internals? Regards, Thomas Hallgren On 03/09/2010 05:47 PM, Thomas M Houser wrote: Pascal, in comment 3 of capilano work item 23579, you said: 3. Pascal Rapicault (archived), Feb 1, 2010, 7:13 PM Only the following things will make it in the API: org.eclipse.equinox.internal.p2.metadata. IRequiredCapability org.eclipse.equinox.internal.provisional.p2. director.IDirector org.eclipse.equinox.internal.provisional.p2. director.IPlanner org.eclipse.equinox.internal.provisional.p2. director.PlanExecutionHelper org.eclipse.equinox.internal.provisional.p2. director.ProfileChangeRequest org.eclipse.equinox.internal.provisional.p2. metadata.MetadataFactory What's up with that? Can you please provide detailed instructions on implementing an equivalent to isOptional() using getMin() and getMax()? The javadoc present in the IRequirement class is just too overwhelming (I'm being sarcastic :-). Also, if it's really true that IRequiredCapability will not be API in 3.6, can you also provide instructions on implementing equivalents to getName(), getNamespace(), and getRange() using IRequirement API.
This got replaced by the getMin getMax on IRequirement. You will also notice that IRequiredCapability is no longer API. On 2010-03-08, at 9:10 PM, Thomas M Houser wrote: In "eclipse-SDK-I20100217-1031-win32.zip", the isOptional() method of IRequiredCapability was removed. This is a change from 3.6M5. Where did this go? What should be used instead?
Good point. The work on MetadataFactory needs to be done. <Mail Attachment.gif>Thomas M Houser ---22/01/2010 04:02:42 PM---I understand that there are potentially many forms of IRequirement. That makes sense, but that does not mean that the API prov
I understand that there are potentially many forms of IRequirement. That makes sense, but that does not mean that the API provided by IRequiredCapability is not still useful to clients. They just need to be aware that this isn't the only kind of IRequirement. For example, we create requirements via MetadataFactory.createRequiredCapability(). How do we create concrete requirements without an API?
IRequiredCapability was limiting in what it could express. For example it could not express negation and or'ing, and it could only described dependencies on something that had a namespace, a name and a version and we are striving to express requirements and capabilities on other things (for example BundleExecutionEnvironment). As such, to ensure for API evolution we have turned the too specific IRequiredCapability into an IRequirement. IProvidedCapability is still API, but it is very likely that it will either be turned into something more generic, or be marked internal if we don't have the time to make it API. As for IRequirementChange the reference to IRequiredCapability is something that we have still left to do (https://bugs.eclipse.org/bugs/show_bug.cgi?id=299506) The situation of the API now is that we have tried to move all the types that we needed to see made API in a final package for M5, however we still have work to do to reach the final state for 3.6 (hopefully by M6). If you have feedback, it is very much appreciated, HTH <Mail Attachment.gif>Thomas M Houser ---22/01/2010 03:23:08 PM---I'm migrating from Eclipse 3.5 to Eclipse 3.6 (I20100119-0800). Can someone explain why IRequiredCapability is now internal in
I'm migrating from Eclipse 3.5 to Eclipse 3.6 (I20100119-0800). Can someone explain why IRequiredCapability is now internal in 3.6 when it was provisional API in 3.5? What is the rationale for making this internal? Why is IProvidedCapability considered API when IRequiredCapability is not? Why does the IRequirementChange API have a matches() method whose parameter type is IRequiredCapability an internal API? _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev _______________________________________________ p2-dev mailing list p2-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/p2-dev |
