Re: RFD: Discarded tuple count for SeqScan nodes in EXPLAIN ANALYZE - Mailing list pgsql-hackers

From Greg Stark
Subject Re: RFD: Discarded tuple count for SeqScan nodes in EXPLAIN ANALYZE
Date
Msg-id 4136ffa0905220911jec20a34ha1830e2dd67e8643@mail.gmail.com
Whole thread Raw
In response to Re: RFD: Discarded tuple count for SeqScan nodes in EXPLAIN ANALYZE  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
On Fri, May 22, 2009 at 4:54 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> It doesn't really seem useful enough to justify breaking client-side
> code that looks at EXPLAIN output.

Fwiw at least pgadmin I don't think would be confused by this. These
tool authors aren't enamoured of fragile assumptions and the
maintenance headaches they cause either.

> This sort of ties into the discussions we have periodically about
> allowing EXPLAIN to output XML or some other more-machine-friendly
> data format.  The barrier for adding additional output fields would
> be a lot lower in such a format.

This is still pretty much true if only for the sheer unscalability of
the amount of data being presented for users to sift through. I do
want us to add a ton more instrumentation into the explain plan and
this is only one small addition. If we add number of hard and soft
i/os, time spent in user and system space, etc the result would be
pretty unreadable and they're at least as important as things like
this.

--
greg


pgsql-hackers by date:

Previous
From: Greg Stark
Date:
Subject: Re: Common Table Expressions applied; some issues remain
Next
From: Greg Smith
Date:
Subject: Revisiting default_statistics_target