On 01 Feb 2013 12:04, Stone, Joel wrote:
I am looking for changes to - say - customer records.
Is it reliable to look at the oldest and newest pair of journal
records for a given file/RRN combination?
For example, a customer's favorite color changed today several
times - from GREEN to red to blue to ORANGE.
Is it reliable to only look at the GREEN (oldest journal record)
and the ORANGE (newest journal record) to find what the current
favorite color is?
Or can the rollbacks get me into trouble with this logic?
Rollback is not a concern; just be sure to look at the after-image.
I think I am OK - if ORANGE changed to yellow with a rollback, I
think that the last journal record would still show ORANGE still,
so I should be OK.
Hmmm.. yellow was never set according to the example; green, red,
blue.? But anyhow, if the after-image of the rollback as the last entry
was showing yellow, and the last activity was that rollback, then the
value was set back to yellow; so yellow, not orange, will be shown in
the last R-UR.
Anything I am overlooking?
Seems the uncommitted changes is what are being overlooked. Those
should not be looked at to conclude what is the current value, because
those changes have neither been committed nor rolled back. The C-CM or
C-RB that follows the changes under the "Commit cycle ID" from the prior
C-SC must occur before those R-xx entries are confirmed.
