At 13:44 02.08.2001 -0500, Waldhoff, Rodney wrote:
>> After taking a cursory look at the code I would suggest
>> to stick with NOPLog and Log4jCategoryLog. Moreover, you
>> probably want to delegate the category look up to log4j
>> instead of maintaining your own hashtable/hashmap. Not
>> only will this be faster but also safer. Just my 2c, Ceki
>
>Thanks Ceki.
>
>The second suggestion follows from the first. As long as SimpleLog exists,
>there is some need for the cache of logs in LogSource, otherwise you'll get
>a distinct instance of SimpleLog, even if it has the same name/category.
Right. Having multiple instances of SimpleLog for the same name would be unacceptable.
>(Well, that's not a critical problem, but it seems less efficient,
>especially if we were to add synchronization per log category.) Not caching
>the log instances would exchange memory for a little bit of time, but I'm
>assuming clients will maintain a static or instance variable referencing the
>Log instance and hence the makeNewLogInstance would be invoked quite rarely.
Assuming that makeNewLogInstance would be invoked quite rarely is reasonable. I don't
understand your point about adding synchronization per log category.
>I'm not sure what you mean by "safer", but I you're more in tune with the
>way log4j operates than I am. Would it make sense to not cache the logs
>that are instances of Log4JCategoryLog?
I was about to write that there was no need to cache instances Log4JCategoryLog but
that is not necessarily true because you do not want multiple instances of
Log4JCategoryLog lying around event if they wrap the exact same
org.apache.log4j.Category instance.
>What's not safe about the alternative? I typically store a static final reference to
>the log4j
>Category anyway and that would seem to be more or less equivalent. Is there
>some danger in that practice I'm not seeing?
Yes and no. As long as there is no Category.remove operation you are safe. However, if
and when Category.remove operation is introduced your wrapped category would be a
dangling category living outside the normal hierarchy. As things stand today, there is
no Category.remove operation and I don't think it would be introduced anytime soon.
Regards, Ceki
--
Ceki G�lc� - http://qos.ch
