Re: [mdr-users] custom implementation recycling?

>>>> If so, still it seems using objects works as long as you don't want 
>>>> to actually store it (i.e. as long as endTrans(false) hasn't been 
>>>> called, everything is accessible and even Objects can be put in 
>>>> slots, and recovered correctly...) Is this indeed correct, or are 
>>>> there situations where there will be problems even before 
>>>> endTrans(false) is called?
>>>>
>>>>
>>>>     
>>>
>>> Yes, since the storage may choose to swap to the disk even data that 
>>> have not been commited yet if a cache threashold is reached.
>>> Martin
>>>   
>>
>>
>> Ic... And what if I use the memory-implementation? it seems like in 
>> that case transactions do work with objects, so I guess the caching 
>> problem is a non-issue there too? Or does even the 
>> memory-implementation swap to disk?
>> If so, is there a way to turn this off? I mean, I'm currently using 
>> MDR for querying purposes only, not for persistence, so it would be 
>> nice to be able to ignore these persistence/serialization issues...
>>  
>>
> I think the memory impl. uses serialization to implement rollback 
> (before an object is change in a transaction, its current state is 
> remembered in a byte array using the serialization routines to be able 
> to restore it in case of rollback). But I am not 100% sure about this. 
> Dan, could you please confirm?
> Martin

Yes, I can confirm this. Objects are being serialized to byte arrays and 
these arrays are passed into transaction log and store there for 
eventual rollback.
Dan