Skip to main content
Skip to main content
Edit this page

Float32 | Float64 | BFloat16 Types

Note

If you need accurate calculations, in particular if you work with financial or business data requiring a high precision, you should consider using Decimal instead.

Floating Point Numbers might lead to inaccurate results as illustrated below:

The equivalent types in ClickHouse and in C are given below:

  • Float32 — float.
  • Float64 — double.

Float types in ClickHouse have the following aliases:

  • Float32 — FLOAT, REAL, SINGLE.
  • Float64 — DOUBLE, DOUBLE PRECISION.

When creating tables, numeric parameters for floating point numbers can be set (e.g. FLOAT(12), FLOAT(15, 22), DOUBLE(12), DOUBLE(4, 18)), but ClickHouse ignores them.

Using Floating-point Numbers​

  • Computations with floating-point numbers might produce a rounding error.
  • The result of the calculation depends on the calculation method (the processor type and architecture of the computer system).
  • Floating-point calculations might result in numbers such as infinity (Inf) and "not-a-number" (NaN). This should be taken into account when processing the results of calculations.
  • When parsing floating-point numbers from text, the result might not be the nearest machine-representable number.

NaN and Inf​

In contrast to standard SQL, ClickHouse supports the following categories of floating-point numbers:

  • Inf – Infinity.
  • -Inf — Negative infinity.
  • NaN — Not a number.

See the rules for NaN sorting in the section ORDER BY clause.

BFloat16​

BFloat16 is a 16-bit floating point data type with 8-bit exponent, sign, and 7-bit mantissa. It is useful for machine learning and AI applications.

ClickHouse supports conversions between Float32 and BFloat16 which can be done using the toFloat32() or toBFloat16 functions.

Note

Most other operations are not supported.