Personally, I've yet to see any actual benchmarks whatsoever on this
issue.
I have seen postings from people with access to unpublished benchmarks
that say the idea that there is a performance hit for introspection is
obsolete with late mode JVM's. (We've come a long way, baby;-)
My personal opinion is that removing introspection as a performance
tweak is like unrolling loops. You can trade space for time, but it's
not a difference that makes a difference in commercial applications.
If we were going to look for performance tweaks, the place to start
would be with a profiler that can pinpoint bottlenecks. Often, an
application will spend 90% of its time processing 10% of the code, and
it's usually not the code you would expect [Fowler, Refactoring (Addison
Wesley 2001), pg 69-70]. If the introspection calls are not in that 10%,
then no amount of performance tuning will make a visible difference.
It's been my observation that my own Struts applications perform
surprisingly well, as does every Struts application I've seen.
-- Ted Husted, Husted dot Com, Fairport NY USA.
-- Custom Software ~ Technical Services.
-- Tel +1 716 737-3463
-- http://www.husted.com/about/struts/
David Corbin wrote:
>
> The collection thing was only an analogy. You missed the general point that
> struts has fine performance for many applications as is, and it allows you
> to create great/powerful/flexible/maintainable code. No one would deny that
> there are cases where introspection might destroy performance, but in many
> cases it is non-issue.
>
> Furthermore, there have been articles and discussions on introspection
> performance that indicate it is not nearly as bad performance wise as some
> believe, though personally I remain a little skeptical about that.
>
> David Corbin
