[XML-DEV Mailing List Archive Home]
[By Thread]
[By Date]
Re: [xml-dev] What should TrAX look like? (Was: Re: [xml-dev] Article on JAXP 1.3 "Fast and Easy XML Processing")
Just run pretty much any query over more than toy data, say 1 - 100 MB
instead of 1 KB.
For example, simply run a benchmark driver such as
java -server nux.xom.tests.XQueryBenchmark 1000 3 xom
samples/data/randj.xml samples/xpath/queries1.xml
The only reason I didn't publish the results for XOM's Jaxen XPath is
because I considered them too embarrassing for you. It really escapes
me why you are asking for it. Shall I really attach the catastrophic
numbers to this mail?
Wolfgang.
>> I remember running a very simple "//item" with the XOM CVS XPath toy.
>> Turns out it would have taken a month to run to completion -
>> quadratic time complexity where one would expect O(N). With another
>> engine it ran in less than a second. Similar for other queries.
>
> If you have an actual test case that demonstrates the issue, I suspect
> I can fix it. However, allegations on mailing lists without any actual
> data rarely result in improvements.
>
> --
> Elliotte Rusty Harold elharo@...
> XML in a Nutshell 3rd Edition Just Published!
> http://www.cafeconleche.org/books/xian3/
> http://www.amazon.com/exec/obidos/ISBN=0596007647/cafeaulaitA/ref=nosim
|
