Re: Performance improvement for joins where outer side is unique

On 9 March 2016 at 13:19, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> I do think that the verbosity this adds to the EXPLAIN output is not
> desirable at all, at least not in text mode. Another line for every
> darn join is a pretty high price.

For me it seems like a good idea to give some sort of indication in
EXPLAIN about this, although I'm not wedded to what I have currently
as being the best and neatest way to do this.
I was however under the impression that you thought this was
reasonable [1], so I didn't go seeking out any other way. Did you have
another idea, or are you just proposing to remove it for text EXPLAIN?
Perhaps, if you are then the tests can still exist with a non-text
output.

Also in [1], I disagree with your proposal for being inconsistent with
what's shown with VERBOSE depending on the EXPLAIN output format. If
we were going to do that then I'd rather just switch on verbose for
non-text, but that might make some people unhappy, so I'd rather we
didn't do that either. So how about I remove it from the TEXT output
and just include it in non-text formats when the VERBOSE flag is set?
then change the tests to use... something other than text... YAML
seems most compact.

[1] http://www.postgresql.org/message-id/8907.1440383377@sss.pgh.pa.us

--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services