Yes, that's interesting and unexpected.

Can you please profile, where the time is lost?
In theory, we even could cache the instance when it is a normal scoped
proxy.

But it is interesting for the overall performance, where the time is
lost...

Cheers,
Arne


Am 20.01.13 19:43 schrieb "Romain Manni-Bucau" unter
<[email protected]>:

>Well it doesnt seem to be as efficient as expected. Replacing a plain old
>injection by an instance i expected no more than x3 in execution time but
>was easily x10...and i had less than 5 beans.
>
>Not sure what takes time...thought to cache the whole result in the
>instance but if you have better ideas it is still open for 1.1.8 and 1.2.0
>:)
>Le 20 janv. 2013 19:24, "Arne Limburg" <[email protected]> a
>écrit :
>
>> Hi Romain,
>>
>> It is cached in the InjectionResolver, that should suffice.
>>
>>
>> Cheers,
>> Arne
>>
>> Am 19.01.13 17:44 schrieb "Romain Manni-Bucau" unter
>> <[email protected]>:
>>
>> >Hi guys,
>> >
>> >is there any reason
>> >why org.apache.webbeans.inject.instance.InstanceImpl#resolveBeans() is
>>not
>> >cached in the instance of InstanceImpl (whatever the cache is - a
>>volatile
>> >var, an atomicref or even a synchronized block)?
>> >
>> >typically doing this pattern: if (!instance.isAmbiguous()) { return
>> >instance.get(); } you'll call it twice for nothing (and it can be O(n)
>> >with
>> >n the numbe rof beans)...or just regarding runtime it seems weird no?
>> >
>> >wdyt?
>> >
>> >*Romain Manni-Bucau*
>> >*Twitter: @rmannibucau <https://twitter.com/rmannibucau>*
>> >*Blog:
>> >**http://rmannibucau.wordpress.com/*<http://rmannibucau.wordpress.com/>
>> >*LinkedIn: **http://fr.linkedin.com/in/rmannibucau*
>> >*Github: https://github.com/rmannibucau*
>>
>>

Reply via email to