Ok, the moment after I wrote this, I realized how deep this thing goes. To do 
what I want here would require seriously rewriting DbAdapter. This is much more 
effort than I am willing to invest now. So I am putting it on hold until we are 
ready to start a general modularity discussion.

Andrus


On Nov 11, 2012, at 1:09 PM, Andrus Adamchik <[email protected]> wrote:
>> However, if it is in a separate jar or not is not important.
> 
> Yeah, good point. I guess I'll proceed and attempt to consolidate everything 
> related to schema and metadata manipulation in 'tools'. Kind of like we 
> consolidated project saving/loading/upgrading/validation functions under 
> cayenne-project in 3.1.
> 
> Andrus
> 
> On Nov 10, 2012, at 10:44 PM, Tore Halset <[email protected]> wrote:
> 
>> Hello.
>> 
>> We use the merge package in our web applications during startup to perform 
>> some not-destructive database migration. Like adding fields, expanding 
>> fields and so on. This makes database migration a lot easier for us since 
>> most of it is done automatically. For us, this is more useful than having 
>> merge in the modeler.
>> 
>> However, if it is in a separate jar or not is not important.
>> 
>> Regards,
>> Tore Halset.
>> 
>> On Nov 10, 2012, at 5:14 PM, Andrus Adamchik <[email protected]> wrote:
>> 
>>> I wonder we should move "merge" package to cayenne-tools from 
>>> cayenne-server in 3.2. To me schema manipulation operations are somewhat 
>>> separate from the runtime part. (So DbLoader and DbGenerator are also 
>>> candidates for a similar move, maybe later).
>>> 
>>> "merge" main use seems to be from the Modeler, which imports 
>>> cayenne-tools.jar. And of course cayenne-tools is available via Maven, and 
>>> other Cayenne distribution channels. 
>>> 
>>> Any reason to keep it in the runtime framework?
>>> 
>>> Andrus
>>> 
>>> P.S. I am also looking at much more glaring modularity issues with Cayenne 
>>> (the whole "unpublished" thing that we discussed on many occasions), but 
>>> don't have any proposals yet, so figured we'd start with things more 
>>> confined in scope. 
>> 
>> 
> 
> 

Reply via email to