Thomas Kellerer <spam_eater@gmx.net> wrote:
> Craig Ringer, 19.01.2010 05:37:
>> regress=> insert into test (x) values ('3');
>> INSERT 0 1
>> regress=> insert into test (x) values ('3.0');
>> INSERT 0 1
>> regress=> insert into test (x) values ('3.00');
>> INSERT 0 1
>> regress=> insert into test (x) values ('3.000');
>> INSERT 0 1
>> regress=> select * from test;
>> x
>> -------
>> 3
>> 3.0
>> 3.00
>> 3.000
>> (4 rows)
> My first question: why does anyone pass a numeric value as a
> string ;)
Forget PostgreSQL for just a moment; try this in Java:
import java.math.BigDecimal;
class BigDecimalTests
{
public static void main(String[] args)
{
BigDecimal x;
x = new BigDecimal(3.000);
System.out.println(x.toPlainString());
x = new BigDecimal("3.000");
System.out.println(x.toPlainString());
x = new BigDecimal(1.01);
System.out.println(x.toPlainString());
x = new BigDecimal("1.01");
System.out.println(x.toPlainString());
}
}
For those without Java to play along at home, the result of
compiling and running this are:
3
3.000
1.0100000000000000088817841970012523233890533447265625
1.01
On top of that, in PostgreSQL '3.000' is *not* a character string,
as you are probably assuming. It is treated as type UNKNOWN until
it has to be resolved (similar to the treatment of a NULL literal),
so it can be interpreted as a literal of some other type.
-Kevin