Re: AW: AW: [Objectbridge-developers] [Fwd: [objectbridge - Open Disc ussi on] New repository.xml/dtd proposal.]

[Ivan T. <to...@cr...>]

On Mon, 2001-09-03 at 10:45, Mahler Thomas wrote:
> Hi Ivan,
>=20
> I had a look at your dtd over the weekend I made some minor
> corrections/additions. See attached dtd. I also modified the repository.x=
ml
> accordingly but could not finish it due to lack of time. but you will get=
 an
> impression...
>=20
> > -----Urspr=FCngliche Nachricht-----
> > Von: Ivan Toshkov [mailto:to...@cr...]
> > Gesendet: Freitag, 31. August 2001 16:37
> > An: Mahler Thomas
> > Cc: objectbridge
> > Betreff: Re: AW: [Objectbridge-developers] [Fwd: [objectbridge - Open
> > Discussi on] New repository.xml/dtd proposal.]
> >=20
> >=20
> > Hi all,
> >=20
> > Hopefully, I've changed everything as we discussed, added the missing
> > `size' and `nullable' attributes and made the xml a little bit more
> > readable.
> >=20
> > Please review it, to see if there is something wrong or missing.
> > Especially with the `ref-id' elements, cause I'm not sure=20
> > I've done them
> > right.
> >=20
> > Is it correct that `ReferenceDescriptor/descriptor_ids'=20
> > points to fields
> > in the current class, and for `CollectionDescriptor/descriptor_ids' --
> > to fields in the referenced class?
> >
>=20
> yes that's right. I tried to make this even clearer by calling the ref-id=
s
> "foreignkey". The foreignkey attribute is always in the table on the N-si=
de
> of the 1-N relationship.
>=20
> >=20
> > Also, I see that the method
> > ObjectReferenceDescriptor.getForeignKeyFields() returns a=20
> > list of ids. I
> > can preserve it's semantics, but I wonder if this is realy needed.
> >=20
>=20
> It's possible to have compound primary keys. Thus it's also possible to h=
ave
> compound foreign keys.
Perhaps, I hadn't formulated my question right. Returning a list is ok,
but why should it contains IDs? Why not list of FieldDescriptors?
=20
>=20
> See my changes to your DTD and my comments below
>=20
> <?xml version=3D"1.0" encoding=3D"UTF-8" ?>
>=20
> <!ELEMENT mapping-repository ( jdbc-connection+, table+, class+,
> class-extent* ) >
>=20
> <!ELEMENT jdbc-connection EMPTY >
> <!ATTLIST jdbc-connection id          ID    #REQUIRED
>                           name        CDATA #REQUIRED
>                           driver      CDATA #REQUIRED
>                           protocol    CDATA #REQUIRED
>                           subprotocol CDATA #REQUIRED
>                           dbalias     CDATA #REQUIRED
>                           username    CDATA #IMPLIED
>                           password    CDATA #IMPLIED >
>=20
> <!ELEMENT table ( column+ ) >
> <!ATTLIST table id       ID    #REQUIRED
>                 schema   CDATA #IMPLIED <!-- we had a schema sttribute wi=
th
> the jdbc-connection,=20
>                                              but it has to be assigned to
> the table.-->
>                 name     CDATA #REQUIRED
>                 jdbc-ref IDREF #REQUIRED >
>=20
> <!ELEMENT column EMPTY >
> <!ATTLIST column id      ID    #REQUIRED  <!-- columns need an ID to be
> referenced from field decriptors  -->
>                  name    CDATA #REQUIRED
>                  type ( BIT | TINYINT | SMALLINT | INTEGER | BIGINT | DOU=
BLE
>                         | FLOAT | REAL | NUMERIC | DECIMAL | CHAR | VARCH=
AR
>                         | LONGVARCHAR | DATE | TIME | TIMESTAMP | BINARY
>                         | VARBINARY | LONGVARBINARY) #REQUIRED
>                  primary-key ( yes | no ) "no"
>                  nullable    ( yes | no ) "yes"
>                  indexed     ( yes | no ) "no"
>                  size CDATA #IMPLIED >
>=20
> <!ELEMENT class ( extent*, field+, reference*, collection* ) > <!-- a
> persistence class must=20
>                                                                     have =
at
> least one field descriptor -->
> <!ATTLIST class name                ID    #REQUIRED
>                 table-ref           IDREF #REQUIRED
>                 orderby-field-ref   IDREF #IMPLIED <!-- you ommited the
> orderby atribute. it's a reference=20
>                                                         to a field
> descriptor -->
>                 conversion-strategy CDATA #IMPLIED
>                 proxy               CDATA #IMPLIED >
>=20
> <!ELEMENT extent EMPTY >
> <!ATTLIST extent class-ref IDREF #REQUIRED > <!-- we only allow classes i=
n
> an extend definition that are=20
>                                                   explained somewhere els=
e
> in the repository -->
>=20
> <!ELEMENT field EMPTY >
> <!ATTLIST field id 	ID	#REQUIRED            <!-- we need references
> to fields in foreignkey-elements -->
>                 name 	CDATA 	#REQUIRED
>                 column-ref IDREF #REQUIRED           <!-- reference to
> column element -->
>                 conversion-strategy CDATA #IMPLIED > <!-- I'd like to hav=
e a
> conversion strategy possible=20
>                                                           for attributes =
too
> -->
>=20
> <!ELEMENT reference (foreignkey+) > <!-- renamed the ref-id element -->
I would preffer to call it foreign-key if that's ok? :)
> <!ATTLIST reference name          CDATA        #REQUIRED
>                     class-ref     IDREF        #REQUIRED <!-- reference t=
o a
> class element -->
>                     auto-retrieve ( yes | no ) "yes"
>                     auto-update   ( yes | no ) "no"
>                     auto-delete   ( yes | no ) "no" >
>=20
> <!-- thma: renamed ref-id ->
> <!ELEMENT foreignkey EMPTY >
> <!ATTLIST foreignkey field-ref IDREF #REQUIRED > <!-- reference to a fiel=
d
> descriptor that contains a foreign key-->
>=20
> <!ELEMENT collection ( foreignkey+ ) >  <!-- renamed the ref-id element -=
->
> <!ATTLIST collection name             CDATA     #REQUIRED
>                      item-class-ref   IDREF     #REQUIRED  <!-- this hold=
s a
> reference to the item class description -->                  =20
>                      collection-class CDATA     #IMPLIED
>                      auto-retrieve ( yes | no ) "yes"
>                      auto-update   ( yes | no ) "no"
>                      auto-delete   ( yes | no ) "no" >
>=20
> <!ELEMENT class-extent ( extent+ ) >
> <!ATTLIST class-extent name CDATA #REQUIRED >
>=20
>=20
> I have a question: What is the scope of an XML ID? The whole xml document=
 or
> only the enclosing element?
The scope is the full document.
> As far as I know they have document scope. thus I had to use fully qualif=
ied
> names in the ID fields (for example for fields and columns), which make t=
he
> repository quite verbose and redundant.
>=20
> Do you see any solution to this?
There are perhaps two solutions:

1) Use XML Schema rather than DTD. I think it supports that kind of
refernces.

2) Check the consistency in the code. That is, no ID attribute for
fields, only name. The foriegn-key element will always be a child of
reference element, which has a class-ref attribute.

I'd vote for the second solution, because it is simpler, cleaner and
later can be extended with the first.
>=20
> Thomas
>=20
>=20
>=20
> > --
> > Ivan
> >=20
>=20