Re: timestamp_part() bug?

> I see following in the manual:
> 
> -------------------------------------------------------------------
> The seconds field, including fractional parts, multiplied by
> 1000. Note that this includes full seconds.
>       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>       SELECT EXTRACT(MILLISECONDS FROM TIME '17:12:28.5');
>       Result: 28500
> -------------------------------------------------------------------
> 
> And I see:
> 
> test=# select current_timestamp,extract(milliseconds from current_timestamp);
>           timestamptz          | date_part 
> -------------------------------+-----------
>  2002-02-27 14:45:53.945529+09 |   945.529
> (1 row)
> 
> Apparently there's an inconsistency among manuals, timestamp(tz)_part
> and timetz_part. Does anybody know which one is correct?

As far as I know, allowing MILLISECONDS etc. for the first arugument
of EXTARCT is a PostgreSQL extention and we should decide what to do
by ourselves.

My proposal is fixing timestamp(tz)_part so that it returns "the
seconds field, including fractional parts, multiplied by > 1000. Note
that this includes full seconds" as the manual stats, since this would
keep the consistency and also have the least impact for existing
applications.

Opinion?
--
Tatsuo Ishi