cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Mazzocchi <stef...@apache.org>
Subject Re: Experimental lazy-loading in ECM++
Date Wed, 05 Jan 2005 20:48:00 GMT
Sylvain Wallez wrote:
> Hi all,
> 
> I (once again) did some refactoring in ECM++ to implement a lazy-loading 
> strategy for components. Using this strategy, a component is loaded 
> (including it's class) only when looked up. This heavily reduces 
> Cocoon's initialization time.
> 
> Some quick measurement of the time spent in Cocoon.initialize() and 
> Cocoon.process() for the first request on my laptop give the following 
> numbers (in seconds):
> - standard "load all" strategy : initialize - 13.6 / process - 4.4
> - lazy loading strategy : initialize - 3.3 / process - 3.7

swish, swish

[sound of Stefano polishing his current hero plate with Sylvain's name 
on top]

> The purpose of this mode is to leverage the autocompiling classloader, 
> by making it usable to automatically restart Cocoon whenever an 
> automatically-compiled class is modified. We would then have the 
> type-safety and IDE-friendlyness of Java combined with the hot-reload 
> abilities of scripted languages.
> 
> This lazy mode is not activated by default. To use it, set 
> JAVA_OPTIONS="-Dorg.apache.cocoon.core.LazyMode=true" before starting 
> "cocoon.sh servlet".

Why not? Given the above speedup times, I would suggest we turn this on 
by default, what do you think?

> ECM++ can be made even more lazy, so expect some more speedup soon.

I have the polishing towel ready :-)

-- 
Stefano.


Mime
View raw message