Mailing-List: contact axis-dev-help@ws.apache.org; run by ezmlm
Precedence: bulk
Reply-To: axis-dev@ws.apache.org
Received-SPF: pass (hermes.apache.org: domain of vreddyp@gmail.com designates
 64.233.170.196 as permitted sender)
DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws;
        s=beta; d=gmail.com;
        h=received:message-id:date:from:reply-to:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references;
        b=Vrf7BHs/IposfcV0HMe1/al9FbRR5n3vcTdw6yOjPlAkwiOKZxeOZHhOmTTtqx7Q37ou19Akn8OSw9yBy6FDpmrUKEQm5iV2msN0FRF4FrpVvL5d5erDC9sd5wKiGp+RmPQEfOSLXH7mvABnvhlDAzWrrBXltgxQBbiS3ynOWTQ=
Message-ID: <fadf663105060300015198e4b6@mail.gmail.com>
Date: Fri, 3 Jun 2005 12:31:13 +0530
From: Venkat Reddy <vreddyp@gmail.com>
Reply-To: Venkat Reddy <vreddyp@gmail.com>
To: axis-dev@ws.apache.org
Subject: Re: [Axis2] Need for children API for OMDocument
In-Reply-To: <429e9528.13ffcb3e.1099.6b31SMTPIN_ADDED@mx.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
References: <fadf663105060121557fa36eed@mail.gmail.com>
	 <429e9528.13ffcb3e.1099.6b31SMTPIN_ADDED@mx.gmail.com>

just saw some more issues if we go with new Interface.

The client API uses OMElement.addChild() to build the message. Even if
we try to use setParent() instead of addChild() here, the user has to
typecast the OMElement into OMElementImpl and pass it to setParent,
which is not elegant.

If we add a new method SOAPFactory.createOMParent(), then the entire
client API has to deal with OMParent objects instead of OMElement
objects while building the message, which looks weird.

- Venkat=20


On 6/2/05, Eran Chinthaka <chinthaka@opensource.lk> wrote:
> Didn't I gave my +1 for this. Ok here goes, +1 ;).
>=20
> -- Chinthaka
>=20
> > -----Original Message-----
> > From: Venkat Reddy [mailto:vreddyp@gmail.com]
> > Sent: Thursday, June 02, 2005 10:55 AM
> > To: axis-dev@ws.apache.org
> > Subject: Re: [Axis2] Need for children API for OMDocument
> >
> > Not another class, but OMElementImpl need to implement the new
> > interface and all the child API be moved from OMElement to OMParent.
> > Also note that OMNode.setParent() now receives OMPraent, instead of
> > OMElement.
> >
> > So the result is - OMDocument now implements only child navigation API
> > and avoids stuff like namaespaces and attributes. +1.
> >
> > - venkat
> >
> > On 6/2/05, Eran Chinthaka <chinthaka@opensource.lk> wrote:
> > > I'm ok with having an "interface" for parent, but not another class.
> > >
> > > -- Chinthaka
> > >
> > > > -----Original Message-----
> > > > From: jayachandra [mailto:jayachandra@gmail.com]
> > > > Sent: Wednesday, June 01, 2005 7:32 PM
> > > > To: axis-dev@ws.apache.org
> > > > Subject: Re: [Axis2] Need for children API for OMDocument
> > > >
> > > > While duplicating the child API into OMDocument I got stuck at
> > something.
> > > > The addChild() method of in turn tries to setParent(), and the
> > > > datamember parent is rigidly typed to be an OMElement only.
> > > >      /**
> > > >      * Field parent
> > > >      */
> > > >     protected OMElementImpl parent;
> > > >
> > > > Now that OMDocument can also be a parent other than just OMElement,=
 my
> > > > suggestion would be to have a wrapper interface OMParent that conta=
ins
> > > > in it the child API methods ( 6 of them). Its good to have child
> > > > navigation API within OMParent than anywhere else (currently it's i=
n
> > > > OMElement). Subsequently OMElementImpl class and OMDocument class i=
f
> > > > they implement this OMParent all the existing code will remain to b=
e
> > > > intact with the additional desired functionality that OMDocument ca=
n
> > > > hold multiple entities in it.
> > > >
> > > > Thanks
> > > > Jaya
> > > >
> > > > My idea boils down to something like
> > > >
> > > > //child navigation API methods will be shifted from OMElement to
> > > > OMParent interface
> > > > public interface OMParent {
> > > >   public void addChild(OMNode omNode);
> > > >   public Iterator getChildrenWithName(QName elementQName) throws
> > > > OMException;
> > > >   public OMElement getFirstChildWithName(QName elementQName) throws
> > > > OMException;
> > > >   public Iterator getChildren();
> > > >   public void setFirstChild(OMNode node);
> > > >   public OMNode getFirstChild();
> > > > }
> > > >
> > > > //OMElementImpl should implement OMParent
> > > > public class OMElementImpl extends OMNodeImpl implements
> > OMParent,...{...}
> > > >
> > > > //OMDocument should implement OMParent
> > > > public class OMDocument implements OMParent {...}
> > > >
> > > > //The parent datamember in OMNodeImpl will be typed as OMParent typ=
e
> > > > public class OMNodeImpl implements OMNode {
> > > > ...
> > > > protected OMParent parent;  // << parent should no longer be
> > OMElementImpl
> > > > type
> > > > ...
> > > > ...
> > > > }
> > > >
> > > > Thank you
> > > > Jayachandra
> > > >
> > > > On 6/1/05, jayachandra <jayachandra@gmail.com> wrote:
> > > > > Yeah! that's a wise way rather than extending OMElement. Apart be=
ing
> > > > > more clear on the readability front it also reduces unnecessary
> > > > > placeholders from creeping into OMDocument.
> > > > >
> > > > > Jaya
> > > > >
> > > > > On 6/1/05, Aleksander Slominski <aslom@cs.indiana.edu> wrote:
> > > > > > jayachandra wrote:
> > > > > >
> > > > > > >Can someone respond on this, plz.
> > > > > > >
> > > > > > >
> > > > > > why not model it exactly as it is in XML Infoset - so have
> > children
> > > > API
> > > > > > but do not extend OMElement just duplicate it (AFAICT Document =
is
> > not
> > > > > > Element ...)
> > > > > > http://www.w3.org/TR/xml-infoset/#infoitem.document
> > > > > >
> > > > > > alek
> > > > > >
> > > > > > >Thanks
> > > > > > >Jaya
> > > > > > >
> > > > > > >On 5/31/05, jayachandra <jayachandra@gmail.com> wrote:
> > > > > > >
> > > > > > >
> > > > > > >>Hi devs,
> > > > > > >>
> > > > > > >>I have two suggestions regarding OMDocument
> > > > > > >>
> > > > > > >>First - a trivial one:
> > > > > > >>---------------------------
> > > > > > >>It lacks an interface definition in the package
> > org.apache.axis.om
> > > > and
> > > > > > >>a direct implementation class with name OMDocument.java is co=
ded
> > in
> > > > > > >>the o.a.a.om.impl.llom package. In line with how rest of the
> > code is
> > > > > > >>arranged, I suggest we have in o.a.a.om package an interface
> > with
> > > > name
> > > > > > >>OMDocument.java listing out the setter and getter methods for
> > > > > > >>rootElement. And in the OMFactory interface we will add an ex=
tra
> > > > > > >>signature something like createOMDocument so as to enable oth=
er
> > than
> > > > > > >>llom factory to be able to provide OMDocument implementation.
> > Let
> > > > the
> > > > > > >>implementation class in impl.llom package be named as
> > > > > > >>OMDocumentImpl.java
> > > > > > >>
> > > > > > >>Second - this is a critical design issue:
> > > > > > >>--------------------------------------------------------
> > > > > > >>Looking at the current OMDocument support I've realized that =
it
> > > > > > >>doesn't have a child navigation API. We might be doing away
> > without
> > > > it
> > > > > > >>as far as soap processing is considered. But without the chil=
d
> > > > > > >>navigation API in it, XMLInfoset can never be fully supported
> > > > because
> > > > > > >>in an XML document other than the unique root element, at the
> > same
> > > > > > >>level we can have several other nodes like documentation
> > comments,
> > > > > > >>processing instructions, DTD element etc.
> > > > > > >>Enabling child API in OMDocument, implementation wise is not =
any
> > > > > > >>difficult. It can be just making it extend OMElement. Somethi=
ng
> > like
> > > > > > >>public interface OMDocument extends OMElement ;
> > > > > > >>
> > > > > > >>Semantically if the above looks confusing and weird (OMDocume=
nt
> > > > being
> > > > > > >>an OMElement !!??!!), alternatively we can copy paste the
> > already
> > > > > > >>coded child API functionality of OMElementImpl into
> > OMDocumentImpl
> > > > > > >>letting OMDocument to stand on its own without extending any
> > other
> > > > > > >>interface. Also, performance wise these changes are not going=
 to
> > add
> > > > > > >>any significant overhead.
> > > > > > >>
> > > > > > >>Anticipating thoughts, ideas, suggestions
> > > > > > >>
> > > > > > >>Regards
> > > > > > >>Jaya
> > > > > > >>
> > > > > > >>
> > > > > > >>
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > The best way to predict the future is to invent it - Alan Kay
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > -- Jaya
> > > > >
> > > >
> > > >
> > > > --
> > > > -- Jaya
> > >
> > >
> > >
> > >
>=20
>=20
>=20
>