CockroachDB supports various inexact, floating-point number data types with up to 17 digits of decimal precision.
They are handled internally using the standard double-precision (64-bit binary-encoded) IEEE754 format.
Names and Aliases
| Name | Aliases |
|---|---|
FLOAT |
None |
REAL |
FLOAT4 |
DOUBLE PRECISION |
FLOAT8 |
Syntax
A constant value of type FLOAT can be entered as a numeric literal.
For example: 1.414 or -1234.
The special IEEE754 values for positive infinity, negative infinity and NaN (Not-a-Number) cannot be entered using numeric literals directly and must be converted using an interpreted literal or an explicit conversion from a string literal instead.
The following values are recognized:
| Syntax | Value |
|---|---|
inf, infinity, +inf, +infinity |
+∞ |
-inf, -infinity |
-∞ |
nan |
NaN (Not-a-Number) |
For example:
FLOAT '+Inf''-Inf'::FLOATCAST('NaN' AS FLOAT)
Size
A FLOAT column supports values up to 8 bytes in width, but the total storage size is likely to be larger due to CockroachDB metadata.
Examples
> CREATE TABLE floats (a FLOAT PRIMARY KEY, b REAL, c DOUBLE PRECISION);
> SHOW COLUMNS FROM floats;
column_name | data_type | is_nullable | column_default | generation_expression | indices | is_hidden
--------------+-----------+-------------+----------------+-----------------------+-----------+------------
a | FLOAT8 | false | NULL | | {primary} | false
b | FLOAT4 | true | NULL | | {primary} | false
c | FLOAT8 | true | NULL | | {primary} | false
(3 rows)
> INSERT INTO floats VALUES (1.012345678901, 2.01234567890123456789, CAST('+Inf' AS FLOAT));
> SELECT * FROM floats;
+----------------+--------------------+------+
| a | b | c |
+----------------+--------------------+------+
| 1.012345678901 | 2.0123456789012346 | +Inf |
+----------------+--------------------+------+
(1 row)
# Note that the value in "b" has been limited to 17 digits.
Supported casting and conversion
FLOAT values can be cast to any of the following data types:
| Type | Details |
|---|---|
INT |
Truncates decimal precision and requires values to be between -2^63 and 2^63-1 |
DECIMAL |
Causes an error to be reported if the value is NaN or +/- Inf. |
BOOL |
0 converts to false; all other values convert to true |
STRING |
-- |
