Mailbox-Line: From andre@liquidprivacy.net  Mon May 20 18:10:20 2002
Content-Type: text/plain;
  charset="iso-8859-1"
From: andre <andre@liquidprivacy.net>
To: pgsql-sql@postgresql.org
Subject: Guarantee uniqueness of PGP key
Date: Mon, 20 May 2002 15:17:09 -0500
Cc: pgsql-novice@postgresql.org
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Message-Id: <20020520201709.A3F8A80269@pepe.lfcity.org>
SPAM: -------------------- Start SpamAssassin results ----------------------
	SPAM: This mail is probably spam. The original message has been altered
	SPAM: so you can recognise or block similar unwanted mail in future.
	SPAM: See http://spamassassin.org/tag/ for more details. SPAM:
	SPAM: Content analysis details:   (5 hits, 5 required)
	SPAM: Hit! (2.0 points)  Received via a relay in relays.osirusoft.com
	SPAM: [RBL check: found 144.42.40.196.relays.osirusoft.com.,
	type: 127.0.0.4]
	SPAM: Hit! (3.0 points)  DNSBL: sender is Confirmed Spam Source SPAM:
	SPAM: -------------------- End of SpamAssassin results
	---------------------

Hi,

More generically, how do I best guarantee that any large varchar is unique=
=20
before inserting?

In my application, I need to ensure that each e-mail address and=20
corresponding PGP public key are both unique.

I originally implemented this by adding a uniqueness contrainst to both=20
columns.  easy.

The problem is that some PGP keys are over 2713 bytes, and postgres (7.2)=
=20
gives the error:

ERROR: btree: index item size 4336 exceeds maximum 2713

I thought perhaps I could use a hash or rtree index, but neither of them wo=
rk=20
for unique-ness.

I saw in another posting that it may be possible to raise this limit to 8k =
or=20
so, but I don't want to do this because 1) it requires a recompile of=20
postgres (and rollout to dev, stage, prod machines), and 2) a key could=20
theoretically be longer than 8k anyway.

I can guarantee unique-ness by removing the index and doing a select before=
=20
the insert.  If there is a matching row, then I am trying to insert a dup.=
=20
bail.

My concern is that doing this without an index will be monumentally slow fo=
r=20
any large-ish number of keys.

Any bright ideas? All suggestions appreciated.

/andre