buildr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michel Guymon <michael.guy...@gmail.com>
Subject Re: ApacheCon / Thoughtworks TechRadar
Date Mon, 12 Nov 2012 17:59:25 GMT
Heya Tammo,

So the goal LockJar is to provide a list of transitive dependencies that 
is consistent over time. This is why the dependency information is 
stored in the Jarfile.lock. Every dependency for a project should be 
resolved to the Jarfile.lock. This prevents conflicts that can appear 
from ad hoc resolution of dependencies when their transitive 
dependencies overlap.

LockJar already has the concept of "group". There is always the default 
group, but you can add any additional group you want. All of LockJar's 
actions (list,load,etc) can handle group.

The perk of the LockJar integration is it automatically populates the 
compile and test task classpaths. That being said, the integration with 
LockJar does not have to be used. I am a bit basis, but I like seeing 
the dependency information in the buildfile. You can actually skip the 
locking and have LockJar resolve every time using something along the 
lines of:

    dependencies = LockJar.list ( :resolve => true ) do
       jar 'org.eclipse.jetty:example-jetty-embedded:jar:8.1.2.v20120308'
    end
    ....
    compile.with project('api'), dependencies

Or if you want to use a single Jarfile for all projects, than you could call

    dependencies = LockJar.list(['test'])
    ...
    test.with project('api'), dependencies

So, uh, I should point out I have not updated LockJar's buildr 
integration lately, since I have been in the land RoR. Mostly I have not 
put much time into it, since there does not appear to be a lot of 
interest in it.

thanks,
Michael

On 11/10/2012 06:46 AM, Tammo van Lessen wrote:
> Hi Michel,
>
> I did have a look at lock_jar and I really like the idea. Since I like on
> one hand Mavens magic resolution for bootstrapping a projects, on the other
> hand I also want to have the control over the dependencies. What I usually
> do is running something like
>
> pp transitive(LOGBACK).map {|a| a.to_spec}
>
> and copy the result into my buildfile. lockjar could automate this for me,
> right?
>
> However, it was difficult to me to grok how a real-world multi-project
> buildr file would look like. If I'm not mistaken, I would need to add all
> dependencies separately to the projects using the lock_jar dsl? I'd love if
> there would be a more unobtrusive way like saying
>
> compile.with project('api'), lock(AXIS2, AXIOM, OPENJPA)
> test.with lock(TESTNG)
>
> and lockjar would use this information to create the lockfile
> (automatically if not there, otherwise on demand) and return a resolved set
> of dependencies. The lock method could also take an optional group
> parameter in order to define independent resolution groups. Still open
> would be the question how excludes or pinned versions could be expressed.
>
> Do you think such an integration would useful or feasible?
>
> Thanks,
>    Tammo
>
>
> On Sat, Nov 10, 2012 at 5:13 AM, Michel Guymon <michael.guymon@gmail.com>wrote:
>
>> Hi,
>>
>> I figure I would pile onto the Buildr vs Maven discussion, this will be a
>> little non sequitur.
>>
>> What I have been doing is using Maven for project info, mostly defining
>> dependencies and where to deploy to. This is what Maven is actually good
>> at, project markup. The perk is all the existing tools can continue to work
>> with the Project's POM. Everything else I do is handled by Buildr. Using
>> LockJar[1] to import the deps and deployment info from the POM, I am able
>> to build and deploy my artifacts from Buildr.
>>
>> I find this is the sweet spot, being able to live in a Maven world but
>> still able to build my artifacts the way I need to with Buildr. God help me
>> if I ever have to use and XML build tool again. . .
>>
>> thanks,
>> Michael
>>
>>
>> [1] https://github.com/mguymon/**lock_jar#buildr-integration<https://github.com/mguymon/lock_jar#buildr-integration><
>> https://github.com/mguymon/**lock_jar<https://github.com/mguymon/lock_jar>
>>
>> On 11/09/2012 10:04 PM, Peter Donald wrote:
>>
>>> Huya,
>>>
>>> On Sat, Nov 10, 2012 at 5:21 AM, Tammo van Lessen <tvanlessen@gmail.com>
>>> wrote:
>>>
>>>> First, I got a Buildr talk accepted at ApacheCon EU in Sinsheim, Germany,
>>>> which I held just yesterday.
>>>>
>>> congrats!
>>>
>>>   I got a lot of positive feedback, people seem
>>>> to like the buildr idea although it also seems that Maven improved since
>>>> the inception of Buildr and some things are less worse. The most visible
>>>> argument against buildr was, however, that people are used to use Maven
>>>> and
>>>> that, if they want to make their project accessible for a broad audience,
>>>> they think they'd need to stick to Maven.
>>>>
>>> That does seem to be a common view.
>>>
>>>   I guess Gradle tries to address
>>>> this issue by providing a "gradle-wrapper", which is a small jar file
>>>> along
>>>> a shell script that you can include into your project and that will
>>>> bootstrap a gradle installation automatically. I also figured, that still
>>>> many Java developers don't have rubies at hand and don't know how to
>>>> easily
>>>> install a gem.
>>>>
>>> And one stage, Antoine was working on the "all-in-one" distribution
>>> that essentially bundled a version of jruby with buildr and all it's
>>> dependencies in one easy installer. I wonder if we could work on this
>>> to ease adoption of buildr for the casual user.
>>>
>>> Where I work we use Chef (http://www.opscode.com/chef/) extensively
>>> and they release their tool in "omnibus" editions that are essentially
>>> a complete version of ruby for n-different platforms. They preinstall
>>> the chef gems in the ruby they distribute but they make sure that the
>>> only things that are added to the path are the che executables. I
>>> wonder if this would be a good thing for us to consider?
>>>
>>>   Second, I stumbled upon ThoughtWorks TechRadar [2]. In particular, I
>>>> liked
>>>> the first paragraph of the Tools section ;)
>>>>
>>> It is kinda neat. Possibly the best thing we can do is to increase
>>> awareness ... I think your approach to giving a talk is a good idea.
>>>
>>>
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message