On Wed, 5 Sep 2007, Heikki Linnakangas wrote:
> Peter Kovacs wrote:
>> One last question:
>>
>> How come that Java double, which is an 8 byte size floating point
>> number, cannot fit in PostgreSQL's double precision type, which is
>> equally a floating point type and 8 bytes in size?
>
> Hmm. I'm not sure when the truncation happens. The value might actually
> still be intact when it's converted from text representation to double,
> and loses precision in the conversion from float8 to decimal. It's
> possible to have a floating point value in the database that you can't
> get out in text format without losing precision. Consider this for example:
>
> postgres=# SELECT
> '6.118992224252588'::float8,
> '6.11899222425259'::float8,
> '6.11899222425259'::float8 = '6.118992224252588'::float8;
> float8 | float8 | ?column?
> ------------------+------------------+----------
> 6.11899222425259 | 6.11899222425259 | f
> (1 row)
>
> Both values are rounded to the same value on output, but if you compare
> them, they're actually not equal.
>
Actually you'll see different behavior with "SET extra_float_digits=2".
Is there a reason we shouldn't set this when we create a connection?
Kris Jurka
