Heads up, BDB coming, needed to support fast configuration

[Roger Riggs]

Hi Chris,

Instead of synchronizing the method on the signature the synchronization
should be done internally.  That will give more flexibility in the 
future if the
internal data structure or interface changes without needing to change 
the method signature.

Roger


On 05/11/2012 08:02 AM, Chris Hegarty wrote:
> On 11/05/12 12:45, Paul Sandoz wrote:
>> Hi Chris,
>>
>> In PersistentTreeMap:
>>
>>   131     public synchronized StringAndInt getStringAndInt(String key)
>>
>>
>> Is that mean to be synchronized? other get methods are not.
>
> In the current implementation then, yes. getStringAndInt currently 
> uses String[] and int[] fields to store the retrieved values. These 
> are of course class wide and can only be used by on call at a time ;-(
>
> PersistentTreeMap can be used in two ways, one instance per persistent 
> on disk db shared among multiple threads, or one instance per thread. 
> The former would benefit from local String[] and int[] instances, but 
> the later would suffer. This will ultimately depend on how the fast 
> configuration work uses PersistenTreeMap. For now, I think we should 
> leave it as is, and revisit if necessary.
>
> Thanks,
> -Chris.
>
>> Paul.
>>
>> On May 10, 2012, at 4:26 PM, Chris Hegarty wrote:
>>
>>> Here is there webrev to support building of the bdb repository.
>>> http://cr.openjdk.java.net/~chegar/jigsaw/bdb_webrev.00/webrev/
>>>
>>> -Chris.