Обсуждение: BUG #7658: the result of pg_get_viewdef can NOT execute directly

The following bug has been logged on the website:

Bug reference:      7658
Logged by:          ZouLixian
Email address:      zoulx1982@163.com
PostgreSQL version: 9.2.0
Operating system:   Linux
Description:        =


postgres=3D# =

postgres=3D# select version();
                                                      version               =

                               =

       =

---------------------------------------------------------------------------=
---------------------------------
-------
 PostgreSQL 9.2.0 on x86_64-unknown-linux-gnu, compiled by gcc (GCC) 4.1.2
20080704 (Asianux 3.0 4.1.2-44), =

64-bit
(1 row)

postgres=3D# =

postgres=3D# create table t1(a int, b int);
CREATE TABLE
postgres=3D# create table t2(c int, d int);
CREATE TABLE
postgres=3D# create view v as select * from t1 t where a in (select a from =
t2
t);
CREATE VIEW
postgres=3D# select pg_get_viewdef('v');
                          pg_get_viewdef                          =

------------------------------------------------------------------
 SELECT t.a, t.b FROM t1 t WHERE (t.a IN (SELECT t.a FROM t2 t));
(1 row)

postgres=3D# SELECT t.a, t.b FROM t1 t WHERE (t.a IN (SELECT t.a FROM t2 t)=
);
ERROR:  column t.a does not exist
LINE 1: SELECT t.a, t.b FROM t1 t WHERE (t.a IN (SELECT t.a FROM t2 ...
                                                        ^
postgres=3D# =

postgres=3D#

zoulx1982@163.com writes:
> postgres=# create table t1(a int, b int);
> CREATE TABLE
> postgres=# create table t2(c int, d int);
> CREATE TABLE
> postgres=# create view v as select * from t1 t where a in (select a from t2
> t);
> CREATE VIEW
> postgres=# select pg_get_viewdef('v');
>                           pg_get_viewdef
> ------------------------------------------------------------------
>  SELECT t.a, t.b FROM t1 t WHERE (t.a IN (SELECT t.a FROM t2 t));
> (1 row)

This is fixed in HEAD as a consequence of commit
http://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=11e131854f8231a21613f834c40fe9d046926387
--- I get

SELECT t.a, t.b FROM t1 t WHERE (t.a IN (SELECT t.a FROM t2 t_1));

which is a valid representation of the view.  At the time we felt it was
not worth the risk of back-patching.  I'm still inclined to think that,
since this example doesn't seem to correspond to any real-world use case
(the original query is surely a mistake no?).

            regards, tom lane