Re: upgrading from 7.3.5 to 8.1.5

[Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>]

Frank Bax <fbax(at)sympatico(dot)ca> writes:
> 1) select and pg_dump produce different results. Why does .06 change to 
> .0601 and .07 to .0699?

It's actually the same values --- remember that float4 is only good to
a bit over six decimal digits anyway.  Post-7.3 versions of pg_dump set
"extra_float_digits" to provide a couple of guard digits in the output:

regression=# select 2006.18::real;
 float4  
---------
 2006.18
(1 row)

regression=# set extra_float_digits TO 2;
SET
regression=# select 2006.18::real;
  float4   
-----------
 2006.1801
(1 row)

On machines with properly-written floating point I/O, this ensures you
get back exactly the same float4 bit pattern you had stored before.
But it looks a bit funny, because the last digit is not "accurate"
in decimal terms.

> 2) age() changed from 7.3.5 to 8.1.5?

I see this in the 7.4.7 release notes:
	Make age(timestamptz) do calculation in local timezone not GMT
It looks like the examples you cite are crossing DST boundaries, so
the one-hour difference is correct.  Depending on what you are trying
to accomplish, you might wish to do the calculation in timestamp without
time zone.

			regards, tom lane