TextWriter discussion

[Fernando Padilla <fe...@interdimensions.com>]

hello.

I would like to open up a topic for discussing among the Xindice 
community.

I don't think it should be Xindice's responsibility to maintain a 
proper(correct) XML serializing code.  Attempting to do so will lead 
to improper XML serialization.  And I am willing to wager that the use of 
our own implementation of XML serialization is at the root of the encoding 
problems Xindice continues to face.

So I feel we should get rid of TextWriter, or implement it as a wrapper 
over more established XML serializing codebases, like Xerces ( if we can 
depend on one parser ), or I believe TrAX is supposed to have serializing 
hooks too.


This is just what I think, what do others think??


Fernando Padilla

RE: TextWriter discussion

[David Viner <dv...@yahoo-inc.com>]

there's a xmlserializer that is provided with xerces2
(org.apache.xml.serialize.XMLSerializer).  I'm not sure if it's the best
serializer available, but it certainly handles	all the commented out cases
in org.apache.xindice.xml.TextWriter.  Most likely the static calls to
TextWriter will need to be handled or re-written.

on a related note, i'm working on the storage of the DTD.  i've made some
minor modifications to org.apache.xindice.xml.dom.DOMParser which properly
add the DocumentType element to the DOM tree.  but the documenttype
information disappears when the document is serialized.

my test case just inserts a document.  it appears that the calls in
Collection.putDocument(Key, Document, boolean) that compress the document
(via DOMCompressor.Compress) and recreate the Document object from the bytes
are the culprit.  So, i looked into the DOMCompressor.Compress method.  i
added a debug statement to the switch statement in Compress, but the
Node.DOCUMENT_TYPE_NODE never seems to be called.  I'm a little confused on
this... can either of you provide pointers as to what i'm doing that's
incorrect?

thanks
dave

-----Original Message-----
From: Kimbro Staken [mailto:kstaken@xmldatabases.org]
Sent: Monday, May 06, 2002 5:18 PM
To: xindice-dev@xml.apache.org
Subject: Re: TextWriter discussion



On Monday, May 6, 2002, at 04:12 PM, Fernando Padilla wrote:

>
> hello.
>
> I would like to open up a topic for discussing among the Xindice
> community.
>
> I don't think it should be Xindice's responsibility to maintain a
> proper(correct) XML serializing code.  Attempting to do so will lead
> to improper XML serialization.  And I am willing to wager that the use of
> our own implementation of XML serialization is at the root of the encoding
> problems Xindice continues to face.
>
> So I feel we should get rid of TextWriter, or implement it as a wrapper
> over more established XML serializing codebases, like Xerces ( if we can
> depend on one parser ), or I believe TrAX is supposed to have serializing
> hooks too.
>
>
> This is just what I think, what do others think??

I believe the reason this was originally done was just performance and the
ability to have an easier way to serialize things. I can't claim to know
all the issues, but I definitely prefer to use existing code wherever
possible so if the custom serializer can be replaced it's probably a good
thing.

>
>
> Fernando Padilla
>
>
>
Kimbro Staken
Java and XML Software, Consulting and Writing http://www.xmldatabases.org/
Apache Xindice native XML database http://xml.apache.org/xindice
XML:DB Initiative http://www.xmldb.org

Re: TextWriter discussion

[Kimbro Staken <ks...@xmldatabases.org>]

On Monday, May 6, 2002, at 04:12 PM, Fernando Padilla wrote:

>
> hello.
>
> I would like to open up a topic for discussing among the Xindice
> community.
>
> I don't think it should be Xindice's responsibility to maintain a
> proper(correct) XML serializing code.  Attempting to do so will lead
> to improper XML serialization.  And I am willing to wager that the use of
> our own implementation of XML serialization is at the root of the encoding
> problems Xindice continues to face.
>
> So I feel we should get rid of TextWriter, or implement it as a wrapper
> over more established XML serializing codebases, like Xerces ( if we can
> depend on one parser ), or I believe TrAX is supposed to have serializing
> hooks too.
>
>
> This is just what I think, what do others think??

I believe the reason this was originally done was just performance and the 
ability to have an easier way to serialize things. I can't claim to know 
all the issues, but I definitely prefer to use existing code wherever 
possible so if the custom serializer can be replaced it's probably a good 
thing.

>
>
> Fernando Padilla
>
>
>
Kimbro Staken
Java and XML Software, Consulting and Writing http://www.xmldatabases.org/
Apache Xindice native XML database http://xml.apache.org/xindice
XML:DB Initiative http://www.xmldb.org