Query should have failed, but didn't?

On 03/01/2011 07:50 PM, Royce Ausburn wrote:

Hi all,

Got an odd one.

test=# select version();
version
-------------------------------------------------------------------------------------------------------------------------------------------
PostgreSQL 8.4.5 on i386-apple-darwin, compiled by GCC i686-apple-darwin8-gcc-4.0.1 (GCC) 4.0.1 (Apple Computer, Inc. build 5370), 32-bit
(1 row)
test=# create temp table test1 (something integer); CREATE TABLE
test=# create temp table test2 (id1 integer, charge numeric); CREATE TABLE
test=# select * from test2 where id1 in (select id1 from test1) and charge=0.05;
id1 | charge
-----+--------
(0 rows)

Notice the last query's subselect is selecting "id1" from test1, but there's no such column on id1. Postgres correctly fails when you do:

test=# select id1 from test1;
ERROR: column "id1" does not exist
LINE 1: select id1 from test1;

The plan for the query that I think should have failed:

test=# explain select * from test2 where id1 in (select id1 from test1) and charge=0.05; QUERY PLAN
-----------------------------------------------------------------
Seq Scan on test2 (cost=0.00..24628.45 rows=3 width=36)
Filter: ((charge = 0.05) AND (SubPlan 1))
SubPlan 1
-> Seq Scan on test1 (cost=0.00..34.00 rows=2400 width=0)
(4 rows)

Notice the sub plan doesn't actually cause any filtering.

This produces incorrect results when there's data in the table:

test=# insert into test1 select 5;
INSERT 0 1
test=# insert into test2 select 10, 70;
INSERT 0 1

test=# select * from test2 where id1 in (select id1 from test1) and charge=70;
id1 | charge
-----+--------
10 | 70
(1 row)
test=# explain select * from test2 where id1 in (select id1 from test1) and charge=70;
QUERY PLAN
-----------------------------------------------------------------
Seq Scan on test2 (cost=0.00..24628.45 rows=3 width=36)
Filter: ((charge = 70::numeric) AND (SubPlan 1))
SubPlan 1
-> Seq Scan on test1 (cost=0.00..34.00 rows=2400 width=0)
(4 rows)

Am I correct in thinking this is a bug in pg?

--Royce

--Royce

Chief Engineer @ Inomial
03 9999 3125
0417 954 640

Its getting id1 from the parent table. (test2) You can use fields from the parent table in subselects.

try this and it'll complain:

select * from test2 where id1 in (select junk from test1) and charge=70;

try:

select * from test2 where id1 in (select something from test1 where something = id1) and charge=70;

-Andy

On 02/03/2011, at 2:16 PM, Andy Colson wrote:

Its getting id1 from the parent table. (test2) You can use fields from the parent table in subselects.

try this and it'll complain:

select * from test2 where id1 in (select junk from test1) and charge=70;

Oh! Of course! What a fool. Thanks =)

--Royce

This is not a bug; given your test queries whenever you reference “id1” you
are ALWAYS referencing the column “id1” in table “test2”.

test=# select * from test2 where id1 in (select id1 from test1) and

charge=70;

 id1 | charge 
-----+--------
  10 |     70
(1 row)

Hint: Consider the results of:

SELECT ‘literal’ FROM table1;

AND

SELECT
t1.*,
(SELECT t2.col2 FROM t2 WHERE t2.id = t1.id)
FROM t1;

David J.