Re: issues with range types, btree_gist and constraints

From: Tomas Vondra <tv(at)fuzzy(dot)cz>
To: pgsql-hackers(at)postgresql(dot)org
Subject: Re: issues with range types, btree_gist and constraints
Date: 2013-02-04 21:04:48
Message-ID: 51102270.6050105@fuzzy.cz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

I've managed to further simplify the test-case, and I've verified that
it's reproducible on current 9.2 and 9.3 branches.

This is the necessary table structure:

-----------------------------------------------------------------------
CREATE TABLE test (
id TEXT,
valid TSRANGE NOT NULL DEFAULT tsrange(NULL, NULL),
CONSTRAINT unique_ids EXCLUDE USING GIST (id WITH =, valid WITH &&)
);

CREATE OR REPLACE FUNCTION skip_existing() RETURNS trigger AS $$
DECLARE
v_exists BOOLEAN;
BEGIN

SELECT TRUE INTO v_exists FROM test WHERE id = NEW.id;
IF v_exists THEN
RETURN NULL;
END IF;

RETURN NEW;

END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER skip_existing BEFORE INSERT ON test FOR EACH ROW EXECUTE
PROCEDURE skip_existing();
-----------------------------------------------------------------------

I've been unable to reproduce the bug with just a single text column.

The trigger simply skips existing records without throwing any error.

Now let's insert some data into the table - I'll use the same samples as
in the previous test-case (http://www.fuzzy.cz/tmp/samples.tgz).

test=# copy test(id) from '/tmp/sample-1.csv';
COPY 20001
test=# copy test(id) from '/tmp/sample-2.csv';
COPY 18590
test=# copy test(id) from '/tmp/sample-1.csv';
COPY 25
test=# copy test(id) from '/tmp/sample-2.csv';
COPY 45

The last two results are really suspicious - it means that some record
were inserted "again", so let's verify that:

test=# select id, count(*) from test group by id having count(*) > 1;
id | count
----------------------------------+-------
0aab4791e1e41f62fd8452ae2c854a34 | 2
0aa08441cd4526b972bb3451d9f8e4ea | 2
0ab969a3333342837484ec0f81bf1e03 | 2
0aea0c33c76b1fe5d123b18cec184dc0 | 2
0af75a99b37be6dde08afaa69de36d29 | 2
0af80d2c2931756b897b3ca5a0055820 | 2
... many more ...

On 9.3 the number of duplicates is much lower, and it's not stable - on
one run I get 6, on the very next one I get 2, then 5 and so on. That
leads me to a suspicion that it might be an uninitialized variable or
something like that.

Tomas

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Simon Riggs 2013-02-04 21:06:50 Re: split rm_name and rm_desc out of rmgr.c
Previous Message Alvaro Herrera 2013-02-04 20:57:05 split rm_name and rm_desc out of rmgr.c