Обсуждение: pgAdmin III - bug with reverse engineered SQL with function indexes

Donald Fraser wrote:

> pgAdmin III July 16th Build
> PostgreSQL 7.3.3
>  
> The reverse engineered SQL for indexes that use a function do not work.
> For example I create an index with the following command:
> CREATE UNIQUE INDEX tbl_security_fullname_key
>   ON tbl_security
>   USING btree (get_securityname_4idx(s_umbname, s_name, s_classname, id));
> pgAdmin III produces:
> CREATE UNIQUE INDEX tbl_security_fullname_key
>   ON public.tbl_security
>   USING btree (public.get_securityname_4idx(s_umbname::citext, 
> s_name::citext, s_classname::citext, id::int4) citext_ops);
> The problem exists with the appended data type on the column names as 
> the parameters of the function.
> If I try to execute pgAdmin's version I get the following error message:
> ERROR:  parser: parse error at or near "::" at character 220.
>  
>  


Donald,
that citext_ops is the operator class, which is correct according to the 
7.3.3 doc. Additionally, the parse error is reported at "::". Please try 
to trace this down,  I don't have a 7.3 server running any more. Please 
attach a "select pg_get_viewdef(OID_of_the_Index)" output, so we can see 
what pgsql likes to see.

Regards,
Andreas

> Donald Fraser wrote:
>
> > pgAdmin III July 16th Build
> > PostgreSQL 7.3.3
> >
> > The reverse engineered SQL for indexes that use a function do not work.
> > For example I create an index with the following command:
> > CREATE UNIQUE INDEX tbl_security_fullname_key
> >   ON tbl_security
> >   USING btree (get_securityname_4idx(s_umbname, s_name, s_classname, id));
> > pgAdmin III produces:
> > CREATE UNIQUE INDEX tbl_security_fullname_key
> >   ON public.tbl_security
> >   USING btree (public.get_securityname_4idx(s_umbname::citext,
> > s_name::citext, s_classname::citext, id::int4) citext_ops);
> > The problem exists with the appended data type on the column names as
> > the parameters of the function.
> > If I try to execute pgAdmin's version I get the following error message:
> > ERROR:  parser: parse error at or near "::" at character 220.
> >
> >
>
>
> Donald,
> that citext_ops is the operator class, which is correct according to the
> 7.3.3 doc. Additionally, the parse error is reported at "::". Please try
> to trace this down,  I don't have a 7.3 server running any more. Please
> attach a "select pg_get_viewdef(OID_of_the_Index)" output, so we can see
> what pgsql likes to see.

There's no problem with the operator class.
So going back to my original index statement: I execute:

CREATE UNIQUE INDEX tbl_security_fullname_key ON tbl_security USING btree (get_securityname_4idx(s_umbname, s_name,
s_classname,id))
 

pgAdmin III generates:

CREATE UNIQUE INDEX tbl_security_fullname_key ON public.tbl_security USING btree
(public.get_securityname_4idx(s_umbname::citext,s_name::citext,
 
s_classname::citext, id::int4) citext_ops);

The select pg_get_indexdef(OID_of_the_Index) produces the exact definition that
I used to create it (but without the line feeds):
CREATE UNIQUE INDEX tbl_security_fullname_key ON tbl_security USING btree
(get_securityname_4idx(s_umbname, s_name, s_classname, id))

Basically what pgAdmin III should generate for 7.3.3 is:
CREATE UNIQUE INDEX tbl_security_fullname_key ON public.tbl_security USING btree
(public.get_securityname_4idx(s_umbname,s_name, s_classname, id)
 
citext_ops);

That is it needs to drop the pg_type information on the column parameters to
the function, which are in this case ::citext ... and ::int4

Regards
Donald Fraser