cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stephan Michels <step...@apache.org>
Subject Re: benchmarking Cocoon?
Date Sat, 19 Apr 2003 17:53:41 GMT



On Sat, 19 Apr 2003, Nicola Ken Barozzi wrote:

> Stephan Michels wrote, On 19/04/2003 17.33:
> ...
> >
> > You also use buffers between each component to measure the time, like
> > the ProfilerBlock do.
>
> But it's flawed. As we found out in the last Stammtisch, since we use
> SAX events, parts of the top elements can influence part of the bottom
> ones, since the top elements go through the fulol pipeline before the
> bottom are finished. Buffering this is no more:
>
> SAX EVENTS
>
>   a  -1-> a1 -2-> a2
>   b  -3-> b3 -4-> b4
>   c  -5-> c5 -6-> c6
>
> BUFFERING
>
>   a  /--> a1  /--> a3
>   b  |    b1  |    b3
>   c--/    c1--/    c3

I don't get it. Generators and Transformators produce n-events
begining with startDocument until endDocument.


1.Test Generator
G -> Buffer until endDocument

(n+1).Test n. Transformer
Buffer -> T -> next Buffer until endDocument

N.Test Serializer
Buffer -> S (time = total - sum from 1.-(N-1).Test)


It is equal, if I test a component alone. I know the input,
I start the time with startDocument and stopps the time if
the component sends endDocument.


Hmm, but yes except, if the components changes the objectmodel
during producing the events, and another depends on it. This is the only
case, in which the test doesn't work. But in the case, the
concept is flawed.

Stephan Michels.


Mime
View raw message