Hi,
I think I have problem with pg transactions ( or I do not understend how
it work).
I have two connections ( running on java clients via jdbc).
I have one simple table tableA (ID int, value float8 ) with one record
(1,1000).
And make this steps:
Connection1-> begin transaction
Connection1-> read the record from tableA ( select values from tableA
where ID = 1). It returns 1000.
Connection1-> increase the value to 1001.
Connection1-> update the new value in tableA by update tableA set value
= 1001 where ID = 1.
Connection2-> begin transaction
Connection2-> read the record from tableA ( select values from tableA
where ID = 1). It returns 1000. !!!
Connection2-> increase the value with 1. It gives 1001.
Connection2-> update the new value in tableA by update tableA set value
= 1001 where ID = 1.
Connection2->commit.
do some other jobs with connection1.
Connection1->commit
Actualy I use the table as a simple counter for unique values and I
expect that Connection2 will wait until Connection1 finish the task.
Is it normal for pg to work so with transactions?
If yes how can I lock all the tables after beginning the transaction?
If no where can I make mistake?
The same example is working well on oracle and all the task is to port
one oracle based application to pg.
Many tanks and regards,
ivan.
