Re: Disk buffering of resultsets

From: Craig Ringer <craig(at)2ndquadrant(dot)com>
To: "Lussier, Denis" <denisl(at)openscg(dot)com>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: John R Pierce <pierce(at)hogranch(dot)com>, List <pgsql-jdbc(at)postgresql(dot)org>
Subject: Re: Disk buffering of resultsets
Date: 2014-10-04 14:44:36
Message-ID: 543007D4.6070109@2ndquadrant.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-jdbc

On 09/23/2014 09:55 AM, Lussier, Denis wrote:
> Wow... really glad Tom chimed in on this. I've been promoting/using
> PG as an enterprise-class database for over a decade and I was
> struggling with the "fact" that the server doesn't iterate thru a cursor
> without bringing it all into memory.

Well, that assertion would've failed the common-sense sanity test anyway.

You can `SELECT * FROM my_100GB_table` on a machine with 1GB of RAM.
Clearly this is impossible if Pg must marshal all the results into RAM
first.

I think John's misapprehension probably stemmed from the fact that libpq
and many other clients *default* to fetching the whole result from the
server into RAM before reporting success to the caller. This makes it
*seem* like the server must marshal the whole result in memory, but it's
the client doing that, not the server.

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

In response to

Browse pgsql-jdbc by date

  From Date Subject
Next Message Craig Ringer 2014-10-04 14:56:38 Re: Disk buffering of resultsets
Previous Message Craig Ringer 2014-10-04 14:42:13 Re: Disk buffering of resultsets