On Wed, Feb 6, 2013 at 11:19 AM, Wells Oliver <wellsoliver@xxxxxxxxx> wrote: > I don't mean to hog my own thread, but the more I look at the hstore type, > the more reasonable it seems. The table is just a serial, a timestamp, and > two columns 'old' and 'new'. The trigger function inserts these values using > hstore(OLD) and hstore(NEW). > > Then, you can select old, new, and new - old, which returns an hstore of > what changed, or you could store this as a third column called 'delta' or > whatever. The hstore of course can be cast to a record, or any other > suitable object. > > Of course, you could not store old and new, and only the delta if you > preferred, but it's nice to have both records. > > Will anyone tell me there's some terrible side effect of this approach that > I am not realizing? If you want hstore-based history logs, you should read through this, either to implement it, or for ideas on how to do your own: http://wiki.postgresql.org/wiki/Audit_trigger_91plus Cheers, Jeff -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
