Yeah, I did make one abortive attempt to implement this at one point, but I
sort of lost my way and stopped before things got *really* ugly. Since then
I've been waiting until I could see the solution clearly in my mind.

-j

On 11/20/06, Jason van Zyl <[EMAIL PROTECTED]> wrote:


On 20 Nov 06, at 3:36 PM 20 Nov 06, John Casey wrote:

> It surely can be. I didn't want to have an overly aggressive design
> goal
> with this initially; my only motivation was to provide easier
> access to
> maven-artifact*.
>

Yes, I think most people looking at it would agree it's a little
complicated and does need to be simplified.

> What I'd ACTUALLY love to do is create/expand a DAG during the POM
> walking
> that we do via the MavenMetadataSource currently, then go back and
> attach
> the resolved artifacts to the nodes of that DAG when we do the real
> resolution.

maven-artifact must have the complete graph in order to be useful.
What's currently doing, I strongly assert, is not viable. In order to
do any meaningful analysis requires the whole graph. I know we have a
listener and you can build it but the graph is a first class citizen
and is required as the strategies we develop must be applied to the
graph not to the peephole we're using now.

> That would give us the ability to reconstruct any subset of the
> dependency graph without re-resolving, and would colocate POM metadata
> alongside the artifact (which is also a huge PITA to recover at
> times).

Absolutely, you could decorate each node with any useful information.
Jung has a very nifty model for this. They use it for visualization
but the same notion would apply for us. Part of fixing many of the
issues with maven-artifact I feel will be resolve by putting a real
graph back into the fix.

> Obviously, this API is nowhere near that sophisticated...if we were to
> implement something like that in maven itself, and really take the
> time to
> see it from the plugin-developer's (AND external component-
> developer's)
> point of view, I'd have no problem scrapping this library in favor
> of that.
>

I would just suggest trying to incorporate your ideas into maven-
artifact sooner rather then later or it will become impossible. I
think what you want is necessary for the long-term viability of maven-
artifact. The graph must be there intact for analysis.

> -john
>
> On 11/20/06, Jason van Zyl <[EMAIL PROTECTED]> wrote:
>>
>>
>> On 20 Nov 06, at 2:01 PM 20 Nov 06, John Casey wrote:
>>
>> > It's meant to be a general-case API for using maven-artifact from
>> > other
>> > systems. The use case I've been working on is allowing
>> Buckminster to
>> > resolve components using Maven's artifact apis.
>> >
>>
>> Is it an experiment the result of which will be rolled back into
>> maven-artifact?
>>
>> jason.
>>
>> > -j
>> >
>> > On 11/18/06, Jason van Zyl <[EMAIL PROTECTED]> wrote:
>> >>
>> >>
>> >> On 10 Nov 06, at 2:16 PM 10 Nov 06, [EMAIL PROTECTED] wrote:
>> >>
>> >> > Author: jdcasey
>> >> > Date: Fri Nov 10 11:16:33 2006
>> >> > New Revision: 473433
>> >> >
>> >> > URL: http://svn.apache.org/viewvc?view=rev&rev=473433
>> >> > Log:
>> >> > Updates to support buckminster maven integration.
>> >> >
>> >>
>> >> Cool, what does this entail. Being able to use Buckminster
>> providers
>> >> in Maven? Or Maven Artifact stuff in Buckminster?
>> >>
>> >> Jason.
>> >>
>> >> >
>> >> >
>> >> >
>> >> >
>> >>
>> >>
>> >>
>> ---------------------------------------------------------------------
>> >> To unsubscribe, e-mail: [EMAIL PROTECTED]
>> >> For additional commands, e-mail: [EMAIL PROTECTED]
>> >>
>> >>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [EMAIL PROTECTED]
>> For additional commands, e-mail: [EMAIL PROTECTED]
>>
>>


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Reply via email to