Re: TimeStamp LAG in editing context

At 10:16 AM 3/6/2003 -0800, you wrote:
>2. The lag only affect what happens when an object is first brought into 
>an editing context.  Regardless of what you do to the lag after this, the 
>object will not be refreshed.

Everyone keeps getting confused about this, so we should be precise with 
our language. You mean "Regardless of what you do to the fetchTimestamp 
after this..." you don't mean "the lag".  An EOEditingContext has a 
fetchTimestamp. This is not a lag time, it's a point-in-time 
timestamp.  There is a static method that uses the word 
'lag'---setDefaultFetchTimestampLag---this controls what the _initial_ 
fetchTimestamp for a newly created EC will be----it will be that many ms 
'lagged' behind the 'now' time.  But an EOEditingContext individually does 
not have a 'lag', it has a fetchTimestamp point-in-time, no lag about it.

Everyone keeps getting confused about this.

>3. Using a low lag time *and* using a refreshing fetch spec caused a lot 
>of problems with decrementSnapshotCountForGlobalID and 
>incrementSnapshotCountForGlobalID.  Someone else said it was OK for them, 
>so this might only affect inheritance.

That sucks. Me, I actually turned off snapshot reference counting 
altogether back in 5.0, because I couldn't figure out what was causing the 
'unable to increment/decrement' exceptions. If someone else figured out 
more of what was causing it, has it been filed as a bug?  One of these days 
I mean to turn it back on and see if it finally works.  There are 
definitely very negative memory consequences to turning off snapshot ref 
counting, like I've done.

--Jonathan


>Chuck
>
>
>Jonathan Rochkind wrote:
>
>>[Un-cross-posted from multiple groups; following up just to eof list].
>>At 10:31 AM 3/6/2003 -0300, Ing.Gaston Blanco wrote:
>>
>>>I´m setting the lag in the editing context to 1 , then i create an EC, 
>>>fetch an object. Then, from a sql terminal i change it, and when i 
>>>refetch the changes are not reflected in my object. I believed that if 
>>>the lag was to small the snapshot would be refreshed, and my object 
>>>updated, but this didn´t happened.
>>
>>Well, here's a stab at explaining what's going on. The 'lag' you've set 
>>refers to seconds (or ms, I forget, but time) _before the creation_ of 
>>the EOEditingContext. If you've set it to 1 second, that means that EC 
>>will make sure not to use snapshots OLDER than 1 second BEFORE the EC was 
>>CREATED.  So changes that occured AFTER the EC was created----these don't 
>>disturb the EC's commitment to not using snapshots older than 1 second 
>>before it was created.
>>Now you could constantly reset the EOEditingContext's fetchTimestamp(). 
>>Note that the static method is called default timestamp LAG and means 
>>that each EC, upon creation, will have it's timestamp set to that much 
>>time earlier. But the instance method that applies to a particular 
>>EOEditingContext is called setFetchTimestamp(), and takes an NSTimestamp 
>>point-in-time---it's not a lag, it's an actual point-in-time, and the EC 
>>will refuse to accept snapshots prior to this point in time.  So you 
>>could be constantly resetting this fetchTimestamp to be just a few 
>>seconds before NOW.  I'm not sure what effect this will have---it could 
>>be deleterious on efficiency, as in-memory EOs are constantly refetched 
>>because they're snapshots are no good. Alternately, it might not even do 
>>what you want: I suspect it only effects _fetches_, it won't effect EOs 
>>in memory, possibly unless you refetch those EOs. But you say you are 
>>doing a refetch, so it might have the consequences you want.  In fact, 
>>this might be a very good idea, but I'm not really sure, it makes me a 
>>little bit nervous.
>>Alternately, you always have available to you: setting 
>>refreshesRefetchedObjects to true on a particular EOFetchSpecification. 
>>This means, regardless of the timestamp on the EC, when you do this 
>>fetch, take the fresh values from the db and use them to create a new 
>>snapshot.  It applies to just that one fetch spec, and the timestamp set 
>>on the EC doesn't matter at all.
>>Hope this helps,
>>--Jonathan
>>
>>>Any clue?????
>>>
>>>thanks
>>>
>>>
>>>Ing. Gaston Blanco
>>>__________________________________________________
>>>
>>>Todavía no tenés tu Ciudad Internet Mail? Obtenelo ahora! - 
>>>http://webmail.ciudad.com.ar
>>>
>>>Descargá Gratis el nuevo Internet Explorer 6.0, el mejor software para 
>>>actualizar tu PC.
>>>http://www.ciudad.com.ar/ar/servicios/ie/
>>>
>>>_______________________________________________
>>>EOF mailing list
>>>[email protected]
>>>http://www.omnigroup.com/mailman/listinfo/eof
>>
>>_______________________________________________
>>EOF mailing list
>>[email protected]
>>http://www.omnigroup.com/mailman/listinfo/eof
>
>
>--
>
>Chuck Hill                                 [email protected]
>Global Village Consulting Inc.             http://www.global-village.net
>
>
>