Re: why table.name is translated to (name.*)::name?

2010/3/30 Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>:
> Hello,
>
> I was noticed on little bit strange feature of PostgreSQL 8.4 and 9.0
>
> we can use a non existing column "name". What does mean?
>
>
> postgres=# create table h(a int, b int);
> CREATE TABLE
> Time: 2,604 ms
> postgres=# insert into h values(199,22);
> INSERT 0 1
> Time: 0,970 ms
> postgres=# explain verbose select h.name from h;
>                         QUERY PLAN
> -------------------------------------------------------------
>  Seq Scan on public.h  (cost=0.00..42.10 rows=2140 width=32)
>   Output: (h.*)::name
> (2 rows)
>
> Time: 0,727 ms
> postgres=# select h.name from h;
>   name
> ----------
>  (199,22)
> (1 row)
>
> Time: 0,589 ms
> postgres=#

FYI this has caused me (and presumably a few other people) a bit of
head-scratching, e.g.:

http://archives.postgresql.org/pgsql-general/2010-03/msg00362.php

I imagine it has some potential as a 'gotcha', as "name" is hardly an
uncommon column name, but it's not an issue which can easily
researched...

Ian Barwick