I've talked to Ken Geis via email. He suggests that there is
considerable overhead to be saved if we go to binary; especially in
date, and timestamp fields
One thing though if the date is 64 bit instead of float, what does
the binary output look like? Are they different ?
If so this would seem to complicate things quite a bit.
Dave
On 14-Nov-05, at 12:12 AM, Tom Lane wrote:
Dave Cramer <[EMAIL PROTECTED]> writes:
I was thinking that it may be necessary to issue a describe before
the execute, but I'm thinking now that the driver can only handle
specific types, so anything outside of what it knows about would be
an error anyway.
I gather it's not possible to mix the return format? For example all
known types would be binary, others would be text ? At this point I'm
not even sure it would help.
You can ask for mixed return formats; see the description of the Bind
message. The sticky spot is that you can't really do that without
first
having gotten the list of output columns (via Describe Statement).
Without that, you don't even know how many output columns there are,
let alone which ones have datatypes you understand.
I'm not sure that this is a fatal objection, at least not for
prepared-in-advance statements. You can put a Describe Statement into
the same network packet exchange as the original Parse message, so
there isn't any reason that you can't know the column types. It is
problematic if you want to Parse/Bind/Execute in just one round trip.
regards, tom lane
---------------------------(end of
broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?
http://archives.postgresql.org
