On Sun, Dec 03, 2006 at 11:24:36PM -0600, John Stanton wrote: > How about running a daemon on your machine which gets the request from > your user defined function, does the lookup with a persistent connection > and asynchronously updates the row in the DB? It does not need to be a > daemon, it could be a thread in your program if that is more apprpriate. > > When it has a few requests answered it can deliver them in one transaction.
Clearly one can live without this feature. It means splitting what could be a single statement/transaction into two at the very least. > If your async process is fed by a FIFO buffer then netwrok slowness and > glitches will not stop your main process. You can optimize traffic by > not sending partially packets, by accumulating requests until you have a > full packet. There is little value in grouping more than that, but > having a persistent connection is valuable. Timeouts while holding on to a DB lock are probably the biggest problem with batcheable functions, yes, and perhaps the best reason to say no to them. Thanks, Nico -- ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
