users@fi.java.net

Re: Non-FI performance problems

From: Tatu Saloranta <cowtowncoder_at_yahoo.com>
Date: Fri, 20 Jan 2006 15:20:54 -0800 (PST)

--- Santiago Pericas-Geertsen
<Santiago.Pericasgeertsen_at_Sun.COM> wrote:

>
> On Jan 19, 2006, at 11:00 AM, Mark Swanson wrote:
>
> > Hello,
> >
> > Before I tested the nice time/space FI compression
> I wanted to make
> > sure that normal non-FI clients would still
> perform well. In
> > summary I found that the FI stax implementation
> for non-FI clients
> > was about 1.85x slower than woodstox and about
> 1.4x slower than the
> > RI.
>
> What do you mean by "FI stax implementation for
> non-FI clients"?
> The FI stax parser can only parse FI, so you must be
> using some other
> parser if the input is XML.
>
> You may also want to take a look at Japex
> (https://
> japex.dev.java.net) where you can find drivers in
> the JDSL (Japex
> drivers standard library) and run performance tests
> for several of
> the mainstream XML parsers --I don't think we have
> one for Woodstox,
> but it should be easy to write by copying one of the
> others.

... and quite naturally, I'd be interested in that
too. ;-)
I have written long-running test suites too (although
not multi-threaded ones that sun's ws group posted;
those were interesting), but haven't had time to look
into japex.

Now, the numbers seen actually are quite in line with
what performance testing I did (for some doc types
like soap messages, and open office docs). More
namespace/attribute heavy documents are, better many
of Woodstox' optimizations work.
Woodstox is quite aggressive in its string sharing,
and that contributes here, as well as use of some
light-weight map implementations for namespace
resolution and attribute value tracking.

-+ Tatu +-


__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com