Re:

Dave Page wrote:
> Oliver reported:
>
> 2. The dump produced:
> CREATE TABLE cust_alloc_history (
> ...
> "year" integer DEFAULT date_part('year'::text,
> ('now'::text)::timestamp(6) with time zone) NOT NULL,
> ...
> ERROR: Column "year" is of type integer but default expression is
> of type double precision
> You will need to rewrite or cast the expression
>
> For an original definition of:
>
> year INTEGER DEFAULT
> date_part('year',CURRENT_TIMESTAMP)

Wow. That is clear. Why are we returning "year" as a double? Yes, I
see now:

test=> \df date_part
List of functions
Result data type | Schema | Name | Argument data types
------------------+------------+-----------+-----------------------------------
double precision | pg_catalog | date_part | text, abstime
double precision | pg_catalog | date_part | text, date
double precision | pg_catalog | date_part | text, interval
double precision | pg_catalog | date_part | text, reltime
double precision | pg_catalog | date_part | text, time with time zone
double precision | pg_catalog | date_part | text, time without time zone
double precision | pg_catalog | date_part | text, timestamp with time zone
double precision | pg_catalog | date_part | text, timestamp without time zone

I would love to say that this is related to change in casts, but that
isn't the case. It is the new double-precision handling of dates; and
I see no easy way to fix this, and you can't fix this after the data
load because the table wasn't created. Yuck.

I have to ask, why are we using a double here rather than a 64-bit
value, if available?

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073