Обсуждение: ExecInitIndexScan: both left and right ops are rel-vars

And here is an old nemesis, thought to have been fixed in 7.0,
reproducible on 7.0beta3 with the following:

DROP TABLE foo;
CREATE TABLE foo (d date);
CREATE UNIQUE INDEX date_uidx ON foo(d);
CREATE UNIQUE INDEX datetime_uidx ON foo(datetime(d));
INSERT INTO foo (d) VALUES ('17-Jun-1995');

DROP TABLE bar;
DROP SEQUENCE bar_id_seq;
CREATE TABLE bar (   id SERIAL,    start_time DATETIME,   duration FLOAT
);
INSERT INTO bar (start_time, duration) VALUES ('17-Jun-1995', 3);

DROP TABLE baz;
DROP SEQUENCE baz_id_seq;
CREATE TABLE baz (   id SERIAL,    bar_id INTEGER,   duration FLOAT
);
INSERT INTO baz (bar_id, duration) SELECT id, duration FROM bar;       
-- Here's the offending query...
SELECT f.d, r.start_time::date, r.duration AS "r_dur", z.duration AS
"z_dur"
FROM foo f, bar r, baz z
WHERE r.id = 2  AND z.bar_id = 2 AND f.d = (r.start_time - '1 day'::interval)::date ;

Ed Loehr wrote:
> 
> And here is an old nemesis, thought to have been fixed in 7.0,
> reproducible on 7.0beta3 with the following:
> 
> DROP TABLE foo;
> CREATE TABLE foo (d date);
> CREATE UNIQUE INDEX date_uidx ON foo(d);
> CREATE UNIQUE INDEX datetime_uidx ON foo(datetime(d));
> INSERT INTO foo (d) VALUES ('17-Jun-1995');
> 
> DROP TABLE bar;
> DROP SEQUENCE bar_id_seq;
> CREATE TABLE bar (
>     id SERIAL,
>     start_time DATETIME,
>     duration FLOAT
> );
> INSERT INTO bar (start_time, duration) VALUES ('17-Jun-1995', 3);
> 
> DROP TABLE baz;
> DROP SEQUENCE baz_id_seq;
> CREATE TABLE baz (
>     id SERIAL,
>     bar_id INTEGER,
>     duration FLOAT
> );
> INSERT INTO baz (bar_id, duration) SELECT id, duration FROM bar;
> 


A final clue:  if I run 'VACUUM ANALYZE' at this point in the script,
before the select, the error disappears.

Regards,
Ed Loehr

> -- Here's the offending query...
> SELECT f.d, r.start_time::date, r.duration AS "r_dur", z.duration AS
> "z_dur"
> FROM foo f, bar r, baz z
> WHERE r.id = 2
>   AND z.bar_id = 2
>   AND f.d = (r.start_time - '1 day'::interval)::date ;

Try it in 7.0.2 --- if it still happens there then I'm interested.
(Don't see it here with current sources.)
        regards, tom lane