incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tim Williams <william...@gmail.com>
Subject Re: hadoop profiles
Date Thu, 27 Nov 2014 01:20:14 GMT
Cool, I should have asked the first time but I'm wanting to get this
in for 0.2.4 - if you'd rather we wait please say something soon:)

--tim

On Wed, Nov 26, 2014 at 12:56 PM, Chris Rohr <rohr.chris@gmail.com> wrote:
> I think I like this also. Thanks Tim!
> On Wed, Nov 26, 2014 at 12:07 PM Andrew <andrew.va@gmail.com> wrote:
>
>> This definitely looks like a good solution. This should also get rid of the
>> maven warning that we always see about using a variable in the version.
>>
>> On Wed, Nov 26, 2014 at 9:51 AM, Aaron McCurry <amccurry@gmail.com> wrote:
>>
>> > I like this method better.  This is how MRUnit handles the various
>> versions
>> > of Hadoop.
>> >
>> > Aaron
>> >
>> > On Wed, Nov 26, 2014 at 9:35 AM, Tim Williams <williamstw@gmail.com>
>> > wrote:
>> >
>> > > We currently use profiles to support building to varying versions of
>> > > Hadoop and, inside the profile, we alter the "projectVersion" tag to
>> > > append the profile name.  This mostly works, but it leads to problems
>> > > publishing snaphot artifacts to some maven repos (e.g. artifactory).
>> > >
>> > > I believe the correct way to get the behavior we want is to leave the
>> > > projectVersion alone and use "classifier" to get the same artifact
>> > > naming behavior.  So, in the profile we'd set some classifier property
>> > > (e.g. <hadoopClassifier>hadoop1</hadoopClassifier>), then in
the jar
>> > > plugin refer to the profile's property (e.g.
>> > > <classifier>${hadoopClassifier}</classifier>) which will result
in the
>> > > classifier string being appended to the artifact name.
>> > >
>> > > The side effect of this is that dependencies on blur would need to
>> > > specify the classifier with their dependency.
>> > >
>> > > Thoughts?
>> > >
>> > > --tim
>> > >
>> > > **NOTE: There's a longer-term "fix" for this that involves introducing
>> > > our own set of interfaces that can be implemented with various hadoop
>> > > "projects" - which would allow us to get rid of profiles all together
>> > > and just use dependencies.  I'm looking for a quicker solution than
>> > > that right now.
>> > >
>> >
>>

Mime
View raw message