Great post. Thanks for the useful info.
One off-topic question - What tool did you use for graphics to visualize three results?
Great post. Thanks for the useful info.
One off-topic question - What tool did you use for graphics to visualize three results?
I'm not convince by this benchmark.
You are not comparing JAXB against stax, because in each 3 tests you're using JAXB and Stax : 1) jaxb uses stax under the cover for reading the stream (There is no such thing as "pure JAXB client") 2) use of default JVM stax implementation 3) use of woodstox stax implementation.
This is just a comparison between two approachs: jaxb on the entire stream, or jaxb on chunks. And it is expected that calling 100,000 times the jaxb unmarshaller is slower than calling it 1 time.
Hi Oleg, I used Excel 2010 graphs.Glad you found the article useful
Seb, you are quite right when you say I don't compare JAXB and STax as technologies, but rather as approaches. My article is not about benchmarking the Java API (for that the excellent engineers at [now] Oracle have already done the job).
My post is entirely about benchmarking the memory consumption and processing speed if one was to use JAXB to unmarshall the entire XML document versus an iterator oriented approach (e.g. with Stax and Woodstox).
The post does not want to give absolute numbers about either JAXB, STax or Woodstox, but rather about the infrastructure expenses that one would encounter when choosing amongst the different approaches.
Apologies if you found the article misleading.
Hi Marco,
I think that Woodstox 4.1 outperforms the others in speed.
Try to run your tests replacing only the Woodstox jars with latest version (4.1)
Thanks for the article.
Can we assume the same conclusion for "Marshalling"? Is there any reason to think that Marshalling a large Java object base into an XML document would behave differently than above?
Excellent article, thank you very much for the work posting this, exactly what I was looking for.
The question which I still have, is about scaling. Which one scales the best?