On Fri, Oct 10, 2014 at 11:38 AM, Peter Geoghegan <pg@heroku.com> wrote:
>> That seems a lot cleaner than the proposal on the Wiki page. If we
>> go that route, it makes sense to fire the BEFORE INSERT triggers
>> before attempting the insert and then fire BEFORE UPDATE triggers
>> before attempting the UPDATE.
By the way, there is no problem with failing to UPDATE, because we
lock rows ahead of the UPDATE. Once a row is locked, the UPDATE cannot
conflict. There is no danger of UPDATE before row-level triggers
firing without then updating (unless the xact aborts, but you know
what I mean). In general, there is no danger of triggers firing more
often than you might consider that they should, with the sole
exception of the fact that we always fire before insert row-level
triggers, even when an UPDATE is the ultimate outcome.
Restarting for conflicts (e.g. handling concurrent insertions with
promise tuples) necessitates a restart, but to the point after before
row-level insert triggers fire, and from a point before any other
triggers have the opportunity to fire.
--
Peter Geoghegan
