[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Query kernel objectsOfClass
On Sun, 15 Apr 2001, Don Berendsen wrote:
> I agree with Falko's earlier comment that querying should not be based on
> objectsOfClass. Using objectsOfClass as the source collection the query
> operates on effectively 'flattens' the object hierarchy.
>
> I thought of two alternatives:
>
> 1. Creating a Query class similar to that defined in JDO.
>
> In this approach the source collection is passed to the Query as a
> parameter. With Ozone this creates a performance problem unless the Query is
> an persistent OzoneObject. Implementing indexes appears problematic.
>
> 2. Defining a Queryable interface that could be implemented in collections
> that would be the source for queries.
>
> The interface would include methods such as 'selectObjects(someFilter)'. The
> interface might also include methods such as 'createIndex(fieldName)' with
> the implementation automaticaly managing the index and utilizing it to
> optimize queries.
>
> With the Queryable interface approach the source collections for queries are
> normally occuring collections in the object hierarchy. For example:
> aCustomer.getInvoices().selectObjects("id = 1234").
>
> The filter example "id = 1234" is not a proposal for the filter format. It's
> just simple to understand at a glance. Note that if one was using
> objectsOfClass for a source collection that the query would have to traverse
> every invoice object in the database.
>
> Comments?
This expresses exactly my ideas regarding the query kernel! :)
selectObjects(filter) and createIndex(collection, attr) IMO is a good core of a
internal query API. Finding a way to specify the filters ("id=1234") might be
tricky but the basic idea is clear.
Falko
>
> Don
>
>
>
>
> >From: Falko Braeutigam <falko@smb-tec.com>
> >Reply-To: ozone-dev@ozone-db.org
> >To: Dasarathi <dasar@tetra.iitm.ernet.in>
> >CC: ozone-dev@ozone-db.org
> >Subject: Re: Query kernel objectsOfClass
> >Date: Sat, 14 Apr 2001 18:00:29 +0200
> >
> >On Sat, 14 Apr 2001, Dasarathi wrote:
> > > hi falko,
> > > ok then
> > > how do u think we should proceed then?
> >
> >I don't know. If you need the objectsOfClass functionality then implement
> >it.
> >Have a look at org.ozoneDB.core.admin. This might give you an idea of how
> >to
> >make your query kernel. The admin interface is provided by AdminImpl. This
> >is
> >a database object. But unlike ordinary database objects it directly
> >interacts
> >with the ozone kernel. This should also work for the query kernel. Well,
> >just
> >an idea.
> >
> >
> >Falko
> >
> > >
> > > das
> > >
> > >
> > >
> > > On Thu, 12 Apr 2001, Falko Braeutigam wrote:
> > >
> > > > On Fri, 13 Apr 2001, Dasarathi wrote:
> > > > > hi falko,
> > > > > we want the query kernel to finally support OQL, SQL etc etc
> > > > > and we thought we will start off with OQL.
> > > > > and we are not searching on this collection of objects,
> > > > > we will mostly do pointer traversal.unless u store names of
> > > > > all objects in a table or something we cant retrieve
> > > > > without objectsofclass. once this gets over then we can think
> > > > > of better ways of doing it.
> > > > > if u think that objectsofclasses is very easy to implement why
> > > > > dont u write that code such that it is used by the kernel alone.
> > > >
> > > > ???
> > > >
> > > > Did I state anywhere that objectsOfClass is very easy to implement? I
> >probably
> > > > lost you here.
> > > >
> > > >
> > > > Falko
> > > >
> > > > >
> > > > > bye
> > > > > das
> > > > >
> > > > >
> > > > > On Thu, 12 Apr 2001, Falko Braeutigam wrote:
> > > > >
> > > > > > On Wed, 11 Apr 2001, Dasarathi wrote:
> > > > > > > The method objectsOfClass in externalDatabase is not
> > > > > > > implemented in ozone-6.1.I need it very much for our
> > > > > > > query kernel work to proceed.
> > > > > > > If someone has done it mail back
> > > > > >
> > > > > > objectsOfClass() was a wrong way in the design of the API and
> >should be finally
> > > > > > removed from the API. If the query kernel needs this function (and
> >surely many
> > > > > > others that are not part of the native ozone API) then it has to
> >implement it
> > > > > > itself. This of course will make the query kernel depend on
> >internal ozone data
> > > > > > structures and APIs but this (in contrast to a ordinary client
> >program) is no
> > > > > > problem.
> > > > > >
> > > > > > Besides, das, it seems that you are following the way to treat all
> >objects of a
> > > > > > class as a collection and apply the search on this class
> >collection. Do you
> > > > > > think this approach is flexible enough to serve as a base for all
> >(at least
> > > > > > many) query fron-ends? I mean, the Java object model itself
> >doesn't have a
> > > > > > possibility to treat all objects of a class as collection. Only
> >SQL/OQL and
> > > > > > related approaches do this...
> > > > > >
> > > > > >
> > > > > > Falko
> > > > > > --
> > > > > >
> >______________________________________________________________________
> > > > > > Falko Braeutigam
> >mailto:falko@smb-tec.com
> > > > > > SMB GmbH
> >http://www.smb-tec.com
> > > > > >
> > > > > >
> > > > --
> > > > ______________________________________________________________________
> > > > Falko Braeutigam mailto:falko@smb-tec.com
> > > > SMB GmbH http://www.smb-tec.com
> > > >
> > > >
> >--
> >______________________________________________________________________
> >Falko Braeutigam mailto:falko@smb-tec.com
> >SMB GmbH http://www.smb-tec.com
> >
>
> _________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
--
______________________________________________________________________
Falko Braeutigam mailto:falko@smb-tec.com
SMB GmbH http://www.smb-tec.com