Reference: Double Type

The types double precision, float, and real are floating-point numbers that are stored in an eight-byte IEEE 754 format.

The double precision type is not suitable for conducting precise calculations, which are essential, for instance, when handling monetary values.

Usage Example

create table constants (
    name text,
    value real
insert into constants
    values ('pi', 3.141592653589793238462643383279502884),
           ('planck', 6.62607015e-34),
           ('avogadro', 6.02214076e23);
select * from constants;
   name   |      value       
 pi       | 3.14159265358979
 planck   |   6.62607015e-34
 avogadro |   6.02214076e+23
(3 rows)

IEEE Special Values

If you need IEEE 754 special values, you need to enter them with explicitly typed literal syntax:

select real 'nan', real 'inf', real '-0';
 ?column? | ?column? | ?column? 
      NaN | Infinity |       -0
(1 row)


Beware of the limited floating-point precision, where rounding errors can quickly add up, and can lead to subtle errors. Consider, e.g., the following query:

with x(i) as (
    values (0.1), 
select sum(i::real) from x;
(1 row)

With exact-precision numeric types, the result of the query would be 0. However, with double precision, the result is only close to zero.

The result also is not stable, i.e., can change indeterministically when repeated: CedarDB executes queries in parallel and thus cannot guarantee the order in which the numbers are added. For the above query, an equally valid result would be:

(1 row)