Improving the performance of psql tab completion

From: Merlin Moncure <mmoncure(at)gmail(dot)com>
To: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Improving the performance of psql tab completion
Date: 2012-10-10 13:33:21
Message-ID: CAHyXU0z-nqatjySsTw3FpMfc6wiFWDb+4iPcMQES7s0KZ4Oz5Q@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hackers,

I have a database with 94059 entries in pg_class. Things are mostly
working fine but psql tab completion is frustratingly slow (around 2.5
seconds on this box). I poked around in psql a bit and saw that the
main culprit was the table visibility condition check. Here's a
typical query (there are other portions unioned in that are not
relevant to performance):

SELECT pg_catalog.quote_ident(c.relname) FROM pg_catalog.pg_class c
WHERE c.relkind IN ('r', 'S', 'v', 'f')
AND substring(pg_catalog.quote_ident(c.relname),1,7)='pg_stat'
AND pg_catalog.pg_table_is_visible(c.oid)
AND c.relnamespace <> (SELECT oid FROM pg_catalog.pg_namespace WHERE
nspname = 'pg_catalog')

By swapping out
AND pg_catalog.pg_table_is_visible(c.oid)

with
AND c.relnamespace in(select oid from pg_namespace where nspname in
(select unnest(current_schemas(true))))

the response time of the tab completion query got knocked down to a
breezy 88ms. Now, this is a bit crude compared to what
RelationIsVisible is doing. In particular, besides checking the schema
path it's doing this:
/*
* If it is in the path, it might still not be
visible; it could be
* hidden by another relation of the same name earlier
in the path. So
* we must do a slow check for conflicting relations.
*/

...but isn't that overkill for tab completion? The simple query above
seems to exhibit the same behavior (for psql) but am I missing
something?

merlin

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2012-10-10 13:45:07 Re: Improving the performance of psql tab completion
Previous Message Marko Kreen 2012-10-10 13:24:20 Re: Successor of MD5 authentication, let's use SCRAM