On 2/8/07, Dennis Cote <[EMAIL PROTECTED]> wrote:
DragonK wrote: > > > Well, using transactions would be a little difficult, because I have a > library which does something like logging to a database. Using > transaction > on a single insert would be useless in my opinion, and on the other > hand I > can't use transactions on the all inserts, since logging is a continuous > process, it doesn't end. > You can basically have a transaction open all the time. Begin a transaction on startup. Insert your log entries as they happen. Execute a commit and another begin after every 100 records, and commit when your program ends. You can also get more sophisticated adding a timer that will also do a commit and begin say one second after the last log entry if you need to. You will get a speedup by nearly a factor of 100, and if you have a power failure or OS crash you will only lose the uncommitted records, not your entire database. HTH Dennis Cote
Thanks for your reply, Dennis - I'll consider this alternative. -- ...it's only a matter of time...