Oliver Jowett wrote:
It is this one:
http://archives.postgresql.org/pgsql-jdbc/2003-05/msg00127.php
Patch summary: use a temporary LO to stream bytea values to the server
without needing to transform them to a potentially large intermediate string
representation (I think).
Correct. You need almost no extra memory, like when dealing with the LO
types. But you will have to VACUUM your pg_largeobject table
periodically, which may be inconvenient.
What impact does this patch have on latency? We do many small transactions
each dealing with small bytea fields, so if this patch ends up doing
additional round-trips per insert, it's going to hurt.
You are absolutely right. Barry has suggested that I change it and only
do it if the data is above a certain length. We set it at 16K. But I
did not implement it because we decided not to add it to the community
driver as the 7.4 backend will provide us with a mechanism for handling
these cases (without a staging area).
From a quick look at your patch, though, it appears to only touch
setBinaryStream()? From memory, we use setBytes(), so this may not be an
issue for us (although we'd like to keep the option of using
setBinaryStream()).
Yes, the idea was use setBinaryStream() if you want to use this method
(we cannot do much for setBytes() anyway as we need the stream). Of
course, Barry's suggestion makes it much better as it allows one to use
setBinaryStream() for small data as well.
If there is a lot of interest in this fix for 7.3 and 7.2 backends, and
you can convince Barry to incorporate it (with his proposed changes) in
the driver, I can ask for time to get it in shape for submission. But I
won't be able to do it unless it goes into the main line of code -- I
already have to work on the 7.4 version (V3 protocol).
Best regards,
Fernando
--
Fernando Nasser
Red Hat Canada Ltd. E-Mail: fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario M4P 2C9
