Re: Modelization: the missing the piece?

From: Andrus Adamchik (andru..bjectstyle.org)
Date: Wed Sep 04 2002 - 13:02:31 EDT

  • Next message: Magic Hat: "Re: Modelization: the missing the piece?"

    [taking this off webobjects-dev to cayenne-devel, since it is no longer
    about WO]

    Igor Clark writes:

    > Hi PA, Andrus,
    >
    > I think PA's "third way" would be of significant help particularly in the
    > type of project which I'm working on, an ASP-model approach. This way you
    > could just define your object model in your own terms and have the
    > framework do the rest of the work for you.

    I guess we can start realistically talking about this feature once we
    implement simpler Cayenne DataObjects, that does not require a special
    superclass. There has been some discussion about it recently:

    http://objectstyle.org/cayenne/lists/cayenne-devel/2002/09/0012.html

     

    > I think it could become a good addition to CayenneModeler but also to
    > Cayenne in general, ie make sure that the API is available
    > programmatically so you can dynamically generate database schema, java
    > code and mappings automatically. Also index generation is something I feel
    > would be a great addition to EOSchemaGeneration & by extension Cayenne...

    Sure. GUI will just call on core API to perform its functions. Just like now
    we have a core implementation of source code generator in
    org.objectstyle.cayenne.gen. This is being called by different front ends -
    GUI and Ant.

    Andrus

     

     

    > Andrus Adamchik wrote:
    >
    >> petite_abeille writes:
    >>
    >>> Hello,
    >>> There is at least three ways to generate a model for the purpose of OR
    >>> mapping:
    >>> (1) Database centric. The database provides all the meta information for
    >>> the model.
    >>> (2) Model centric. The model is painstakingly hand crafted and both
    >>> database and objects are derived from it.
    >>> (3) Object centric. The object structure itself provides all the
    >>> information for generating a model. Model and database are a side effect
    >>> of the object structure.
    >>> (1) and (2) are how EOF deal with models as of today. One need to have a
    >>> model before being able to do anything else. Further more, all the
    >>> helper tools (eg EOModeler) are tailored toward a very database centric
    >>> approach: recovering database meta information to generate a model and
    >>> so on.
    >>> There is nothing to help with the third case: generating a model based
    >>> on the object's meta information.
    >>> But EOF stands for Enterprise *Object* Framework, not for ETF
    >>> (Enterprise Table Framework), isn't it?
    >>> So why ignoring the "third way"?
    >>
    >>
    >> This could become a good addition to CayenneModeler :-).
    >> Andrus
    >> _______________________________________________



    This archive was generated by hypermail 2.0.0 : Wed Sep 04 2002 - 13:02:36 EDT