PostgreSQL REAL and DOUBLE PRECISION Types

PostgreSQL REAL and DOUBLE PRECISION store approximate binary floating-point values. Use them for measurements and scientific calculations when a small rounding difference is acceptable. Use NUMERIC for exact decimal values such as prices, and integer types for whole-number counts.

The table shows positive magnitudes. Corresponding negative values are also representable. Values too large or too small for the type cause an error; a value too close to zero to remain distinct from zero can also cause an underflow error.

FLOAT(p) selects a type by its binary precision p: values from 1 to 24 select REAL, and 25 to 53 select DOUBLE PRECISION. FLOAT without a precision is DOUBLE PRECISION.

Declare a floating-point column

Use REAL when single precision is sufficient, or DOUBLE PRECISION when you need more range and precision:

CREATE TABLE sensor_readings (
    reading_id bigint GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
    temperature real NOT NULL,
    pressure double precision NOT NULL
);

Floating-point values are approximate

Binary floating-point cannot represent many decimal fractions exactly. For example, multiplying a binary approximation of 0.1 by 3 can produce a small visible difference from exact decimal arithmetic:

SELECT
    0.1::numeric * 3 AS exact_decimal,
    0.1::double precision * 3 AS approximate_float;

 exact_decimal | approximate_float
---------------+-------------------
 0.3           | 0.30000000000000004

Avoid relying on exact equality for floating-point calculations. When comparing measured values, choose an application-appropriate tolerance. PostgreSQL REAL and DOUBLE PRECISION use IEEE 754 binary floating-point where supported by the platform.

Further reading

• PostgreSQL numeric types
• PostgreSQL NUMERIC and DECIMAL
• PostgreSQL integer types