2010/1/20 Adrian von Bidder <avbidder@xxxxxxxxxxx>: > Hi, > > [ creating db partitions on demand ] > > On Wednesday 20 January 2010 11.20:21 Vincenzo Romano wrote: >> In case 1 I need to inspect the catalog with at least a select, while >> in case 2 I need to trap errors. >> In my (little) experience trapping errors is slow, so I would go for >> option 1. > > Trapping/handling the error might be slow, but remember that creating a new > partition (presumably) doesn't happen often (and creating the partition is > slow anyway.) In case 1 (check if the partition exists before insert), > *every* *single* *insert* is slower because you first query the catalog, > while in case 2, the normal case (insert into existing partition) is fast > and only the rare case (how is your partitioning? 1000000 or more rows per > partition on average? I'd think it's quite a lot because why else partition > the data at all...?) is slow. > > cheers > -- vbi Hmmm ... also trapping would happen for every single line being inserted ... -- Vincenzo Romano NotOrAnd Information Technologies NON QVIETIS MARIBVS NAVTA PERITVS -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
