Strange behavior on non-existent field in subselect?

[Ken Tanzer <ktanzer@xxxxxxxx>]

We're a little puzzled by this (apparently) strange behavior, and would 
be curious to know what you folks make of it.  Thanks.

Ken

CREATE TABLE foo (
    foo_field   integer );

CREATE TABLE par(
    par_field   integer );

SELECT VERSION();

SELECT foo_field FROM par;
SELECT foo_field FROM foo WHERE foo_field IN (SELECT foo_field FROM par);
INSERT INTO foo VALUES (1);
SELECT foo_field FROM foo WHERE foo_field IN (SELECT foo_field FROM par);
INSERT INTO par VALUES (1);
SELECT foo_field FROM foo WHERE foo_field IN (SELECT foo_field FROM par);
/* One row for every foo record, provided at least one record in par */

Which (for us) yields the following output:

Chasers=> \i strangefield.sql
CREATE TABLE
CREATE TABLE
                                                version
-------------------------------------------------------------------------------------------------------
 PostgreSQL 8.1.4 on i686-redhat-linux-gnu, compiled by GCC gcc (GCC)
3.4.4 20050721 (Red Hat 3.4.4-2)
(1 row)

psql:strangefield.sql:11: ERROR:  column "foo_field" does not exist
 foo_field
-----------
(0 rows)

INSERT 0 1
 foo_field
-----------
(0 rows)

INSERT 0 1
 foo_field
-----------
         1
(1 row)


begin:vcard
fn:Kenneth Tanzer
n:Tanzer;Kenneth
org:Downtown Emergency Service Center;Information Services
adr:;;515 Third Avenue;Seattle;WA;98104;USA
email;internet:ktanzer@xxxxxxxx
title:Director of Information Services
tel;work:(206) 464-1570 x 3061
tel;fax:(206) 624-4196
x-mozilla-html:TRUE
url:http://www.desc.org
version:2.1
end:vcard