Daniel Kulp wrote:
> Raymond,
> 
> 
>>1) What's the use case to save a XMLDocument or DataObject to the
>>XMLStreamWriter? If there's no such requirement, we can get rid of them.
> 
> 
> Well, part of the work we are doing in Celtix is to allow the "databinding" 
> layer to write directly toward a XMLStreamWriter.   I was hoping to use those 
> methods since that is ideal.
> 
> 

That's kind of what I had in mind - support for someone who wanted to
push data into an output stream.

>>
2) With the createStreamReader() method, we can easily pipe the reader to
>>writer with a utility (Axiom has one). Here's the implementation in my
>>latest patch.
>>
>>      XMLStreamReader reader = createXMLStreamReader(document);
>>      new XMLStreamSerializer().serialize(reader, writer); //
>>XMLStreamSerializer is the utility to pull events from the reader and dump
>>them into the writer
> 
> 
> What is the performance impact of doing that?   I'm OK with doing something 
> like that (and maybe even making the mentioned save() and saveObject() 
> methods do that) if there isn't a major performance issue with doing that.
> 
> Basically, the goal is to get the objects written out to a stream as quickly 
> as possible.   If "reading events from one stream and pumping them to the 
> output stream" isn't too bad, I'm OK with it.   But if there is a performance 
> impact of doing that, then I would suggest leaving the ability to write 
> directly to the XMLStreamWriter.
> 

The biggest performance concern I would have is if this implementation
needed to create a Thread to drive the event pump.

A secondary concern would be if this created a lot of objects for the
various read events e.g. does pushing text mean that an additional
String or char[] instance is created?

Finally, is there any downside to providing the API call? It sounds like
it was something that fitted Dan's model.

--
Jeremy

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to