Hi, I reviewed these changes a bit in the context of multiple contributions. I guess things should work even then because these steps happen in the course of 'addContribution'. So if a new contribution comes in, the definitions in that would further get aggregated and then the composites coming in would be parsed and processed against the aggregated union of all definitions. Do you see something missing ?
Thanks - Venkat On Wed, Feb 13, 2008 at 6:45 PM, Simon Laws <[EMAIL PROTECTED]> wrote: > On Feb 12, 2008 5:18 PM, Venkata Krishnan <[EMAIL PROTECTED]> wrote: > > > Hi, > > > > No there isn't a separate phase. Just that in the current read phase I > > look > > for *.composite files and set those aside in a list without processing > > them > > further. After all artifacts in the contribution have been read I then > > read > > the list of composite URIs, read them and modify them with the > additional > > attribute 'applicablePolicySets' and then push it further for the usual > > processing. > > > > I see that this is what you have also summarized on the wiki. I have > > assumed that in the section titled "New Policy Processing Phase" should > go > > the description of what we do now with the composite reading and > > augmenting. I have added that information. Let me know if your > thoughts > > for it were otherwise. > > > > I think I might have to change this a bit in the context of multiple > > contributions. Isn't it ? > > > > - Venkat > > > > On Feb 12, 2008 2:41 PM, Simon Laws <[EMAIL PROTECTED]> wrote: > > > > > snip.. > > > > > > On Feb 12, 2008 8:40 AM, Venkata Krishnan <[EMAIL PROTECTED]> > wrote: > > > > > > > Yes. Because we are now computing the 'applicablePolicySets' for > > > various > > > > SCA artifacts and that needs the list of 'all' PolicySets that might > > be > > > > applicable ever. > > > > > > > > > > > So, in the code today, how do you know you have reached the point that > > all > > > contributions have been added and you can start associating policy > sets > > > with > > > composites? Is the composite processing now in a separate phase > > > independent > > > of the the contribution processing. > > > > > > To try and get this clearer in my mind I've written out a part of the > > > various phases on the wiki [1]. Is there a new phase? Looking at the > > code > > > I > > > don't see it. > > > > > > Simon > > > > > > [1] > > http://cwiki.apache.org/confluence/display/TUSCANYWIKI/Runtime+Phases > > > > > > > Hi Venkat > > Thanks for the updates. The multiple contribution case was the case that I > was thinking about :-) So you have these new steps that have to sit > between > the point where all the definitions.xml files are read from ALL the > contributions and the point at which a composite is parsed and turned into > an assembly model. SO it sounds like the process would be something like > > 1. Add contribution > 2. Process contribution to extract definitions.xml > > repeat 1 & 2 until all contributions are added > > 3. Find composites in contributions > 4. Process "appliesTo" with reference to each of the composites > 5. process the the composites into an assembly model for further domain > processing (the domain composite) > > I'm not necessarily advocating enforcing the approach that all > contributions > must be added before any further processing commences. You could imagine > an > approach where the process is just repeated as new contributions are added > for example. But you get my point. > > Simon >
