On 4/4/06, Nathan Kurz <[EMAIL PROTECTED]> wrote: > > > >> 3. The performance for inserts is really bad. Around 40k entries > takes a > > >> few hours. What might I be doing wrong? I do a commit after > > >> all the inserts. > > > > > > A few things to help with speed: > > > > > > 1. Use DBI's prepared statements; eg, 1 prepare() and many execute(). > > > > Yes, this is what I do. > > > > > > 2. Don't commit for each row inserted but batch them so, say, you > > > commit once per 1000 rows. > > > > > Unfortunately, I cannot commit till I do all the inserts. > > That doesn't seem right for speed. In addition to using "commit", are > you beginning a transaction with "begin"? Are your inserts > particularly complex or large? More details about what you are doing > would probably be good here, since something odd is happening here. > Maybe you could post a tiny test program along with the time it takes? > > --nate > > I don't begin the transaction with begin. My assumption was that the first insert operation would automatically begin a transaction. My inserts are fairly simple with two columsn being long strings of length 255.
my @values = ($task_info_gid,$file_type_gid,$extracted_path,$media_path, $size,$ctime,$mtime,$job_id,$is_in_du); Raj
