Date: Sun, 4 Dec 2016 10:32:37 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: pgsql-general@postgresql.org
Subject: INSERT ON CONFLICT of "wide" table: target lists can have at most
 1664 entries
Message-ID: <20161204163237.GA8030@telsasoft.com>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="WIyZ46R2i8wDzkSu"
Content-Disposition: inline
User-Agent: Mutt/1.5.23 (2014-03-12)
Precedence: bulk
Sender: pgsql-general-owner@postgresql.org


--WIyZ46R2i8wDzkSu
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

Our application INSERTs data from external sources, and infrequently UPDATEs
the previously-inserted data (currently, it first SELECTs to determine whether
to UPDATE).

I'm implementing unique indices to allow "upsert" (and pg_repack and..), but
running into a problem when the table has >830 columns (we have some tables
which are at the 1600 column limit, and have previously worked around that
limit using arrays or multiple tables).

I tried to work around the upsert problem by using pygresql inline=True
(instead of default PREPAREd statements) but both have the same issue.

I created a test script which demonstrates the problem (attached).

It seems to me that there's currently no way to "upsert" such a wide table?

I see:
./src/include/access/htup_details.h:#define MaxTupleAttributeNumber 1664        /* 8 * 208 */

./src/backend/parser/parse_node.c-      /*
./src/backend/parser/parse_node.c-       * Check that we did not produce too many resnos; at the very least we
./src/backend/parser/parse_node.c-       * cannot allow more than 2^16, since that would exceed the range of a
./src/backend/parser/parse_node.c:       * AttrNumber. It seems safest to use MaxTupleAttributeNumber.
./src/backend/parser/parse_node.c-       */
./src/backend/parser/parse_node.c:      if (pstate->p_next_resno - 1 > MaxTupleAttributeNumber)
./src/backend/parser/parse_node.c-              ereport(ERROR,
./src/backend/parser/parse_node.c-                              (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
./src/backend/parser/parse_node.c:                               errmsg("target lists can have at most %d entries",
./src/backend/parser/parse_node.c:                                              MaxTupleAttributeNumber)));

Thanks in advance for any suggestions.

Justin

--WIyZ46R2i8wDzkSu
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="wide_upsert.txt"

#! /bin/sh
set -e

n=831
t=wide_upsert

psql -c "DROP TABLE IF EXISTS $t"

cols='id int'
vals='0'
sets='id=0'
for a in `seq -w 0 $n`
do
	c="c$a int"
	cols="$cols, c$a int"

	vals="$vals, \$1"
	sets="$sets, c$a=\$1"
done

echo $cols
psql -c "CREATE TABLE $t ($cols, PRIMARY KEY (id))"

set -x
psql -c "PREPARE x AS INSERT INTO $t VALUES ($vals) ON CONFLICT (id) DO UPDATE SET $sets; EXECUTE x(0)"

--WIyZ46R2i8wDzkSu
Content-Type: text/plain
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0


-- 
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

--WIyZ46R2i8wDzkSu--