Re: UPSERT

Josh Berkus <josh@agliodbs.com>

From: Josh Berkus <josh@agliodbs.com>
To: pgsql-hackers@postgresql.org
Cc: Tom Lane <tgl@sss.pgh.pa.us>, "Simon Riggs" <simon@2ndquadrant.com>, "Heikki Linnakangas" <heikki@enterprisedb.com>, "Andrew Dunstan" <andrew@dunslane.net>, "Jonathan Scher" <js@oxado.com>
Date: 2007-03-02T19:17:46Z
Lists: pgsql-hackers
Couple notes:

(1) Upsert is not just a desire of MySQL users.  I just spec'd a major 
proprietary-database replacement project at a fortune 500 where they want an 
Upsert and are unhappy that PostgreSQL doesn't have it.  Unfortunately, they 
don't want to spring for development funds :-(

(2) Doing upsert by checking for a unique key violaton error leads to horrible 
performance in addition to the previously mentioned race conditions.

-- 
Josh Berkus
PostgreSQL @ Sun
San Francisco