Kris Jurka wrote:
>On Tue, 2 Nov 2004, Alan Stange wrote:
>
>
>
>>>The real problem is that executeBatch() returns an int[] of update counts.
>>>
>>>
>>>
>>This was what I was thinking. What does the pg74 driver do in this
>>case as it does allow multiple statements in each batch?
>>
>>
>>
>
>The 8.0 driver uses the extended query protocol while the 7.4 driver
>uses the simple query protocol. The relevent difference in this case is
>that you can only send one command per query in the extended query
>protocol while you can string them together with ";" in the simple case.
>This means the 8.0 driver splits the queries and issues them individually
>and gets update counts for each command. The 7.4 driver issues them
>together and only gets one update count (for the last piece of the batch).
>
>
>
>>Network latency. We were able to greatly increase performance this way
>>by reducing the number of round trips.
>>
>>
>>
>
>Well in the 8.0 case you've lost that anyway because it splits them and
>issues them separately. Although the 8.0 driver does have some other
>advantages in that it doesn't issue a Sync after each individual query,
>but sends out a number of commands and then processes a number of results,
>so the number of round trips will be reduced, but in a different fashion.
>
>
What about this case:
Connection conn = ...
Statement st = conn.createStatement();
st.execute("insert a; insert b; insert c; insert d;");
Is the 8.0 driver busting this up into 4 trips to the database?!
-- Alan
