On 07/20/2014 12:55 PM, Peter Geoghegan wrote:
> There is a *really* big
> demand for UPSERT from users, not MERGE, although MERGE is certainly
> useful too.
The inability to efficiently say "Add this unique-keyed row, or if a row
of the same key already exists replace it atomically" is a fundamental
defect in SQL its self. Vendors shouldn't need to be coming up with
their own versions because the standard should really cover this - much
like LIMIT and OFFSET.
It's very high in the most frequently asked questions on Stack Overflow,
right up there with questions about pg_hba.conf, connection issues on OS
X, etc.
I'd be very keen to see atomic upsert in Pg. Please Cc me on any patches
/ discussion, I'll be an eager tester.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services