# `floor` function

Returns the largest number not bigger than `expr` rounded down to `targetScale` digits relative to the decimal point.

## Syntax

```floor(expr [, targetScale])
```

## Arguments

• `expr`: An expression that evaluates to a numeric.

• `targetScale`: An optional INTEGER literal greater than `-38` specifying by how many digits after the decimal points to round down.

Since: Databricks Runtime 10.5

## Returns

If no `targetScale` is given:

• If `expr` is `DECIMAL(p, s)`, returns `DECIMAL(p - s + 1, 0)`.

• For all other cases, returns a BIGINT.

If `targetScale` is specified and `expr` is a:

• `TINYINT`

Returns a `DECIMAL(p, 0)` with `p = max(3, -targetScale + 1)`.

• `SMALLINT`

Returns a `DECIMAL(p, 0)` with `p = max(5, -targetScale + 1)`.

• `INTEGER`

Returns a `DECIMAL(p, 0)` with `p = max(10, -targetScale + 1))`.

• `BIGINT`

Returns a `DECIMAL(p, 0)` with `p = max(20, -targetScale + 1))`.

• `FLOAT`

Returns a `DECIMAL(p, s)` with `p = max(14, -targetScale + 1))` and `s = min(7, max(0, targetScale))`

• `DOUBLE`

Returns a `DECIMAL(p, s)` with `p = max(30, -targetScale + 1))` and `s = min(15, max(0, targetScale))`

• `DECIMAL(p_in, s_in)`

Returns a `DECIMAL(p, s)` with `p = max(p_in - s_in + 1, -targetScale + 1))` and `s = min(s_in, max(0, targetScale))`

If `targetScale` is negative the rounding occurs to `-targetScale` digits to the left of the decimal point.

The default `targetScale` is 0, which rounds down to the next smaller integral number.

## Examples

```> SELECT floor(-0.1);
-1

> SELECT floor(5);
5

> SELECT floor(3345.1, -2);
3300

> SELECT floor(-12.345, 1);
-12.4
```