On 2010-09-23 1:31 PM +0300, Boxuan Zhai wrote:
> I have just generate a new patch of MERGE command.
I haven't followed the discussion very closely, but this part in the
regression tests caught my attention:
+-- we now have a duplicate key in Buy, so when we join to
+-- Stock we will generate 2 matching rows, not one.
+-- According to standard this command should fail.
+-- But it suceeds in PostgreSQL implementation by simply ignoring the
second
It doesn't seem like a very good idea to go against the standard here.
The "second" row is not well defined in this case so the results are
unpredictable.
The patch is also missing a (trivial) change to explain.c.
Regards,
Marko Tiikkaja