Floor¶

Overview¶

The Floor expression computes the largest integer less than or equal to the input value. It extends UnaryMathExpression and uses math.floor as the underlying mathematical function, with optimized handling for decimal and long integer types.

Syntax¶

FLOOR(expr)

// DataFrame API
import org.apache.spark.sql.functions.floor
df.select(floor(col("column_name")))

Arguments¶

Argument	Type	Description
child	Expression	The numeric expression to compute the floor value for

Return Type¶

The return type depends on the input data type:

For DecimalType with scale 0: Returns the same DecimalType
For DecimalType with scale > 0: Returns DecimalType.bounded(precision - scale + 1, 0)
For all other supported types: Returns LongType

Supported Data Types¶

The expression accepts the following input data types:

DoubleType
DecimalType (any precision and scale)
LongType

Algorithm¶

The evaluation algorithm varies by input data type:

For LongType: Returns the input value unchanged (already an integer)
For DoubleType: Applies math.floor() function and converts result to Long
For DecimalType: Uses the Decimal class's built-in floor() method
Null inputs are handled by the parent class and return null results

Partitioning Behavior¶

This expression preserves partitioning characteristics:

Does not require data shuffling
Can be applied within each partition independently
Maintains existing data distribution patterns

Edge Cases¶

Null handling: Null inputs return null outputs (handled by UnaryMathExpression parent class)
Long integer inputs: Values are returned unchanged since they are already integers
Decimal precision: For decimals with scale 0, the original type is preserved
Decimal scale reduction: For decimals with scale > 0, the result has scale 0 with adjusted precision
Double overflow: Large double values may overflow when converted to Long type

Code Generation¶

This expression supports Tungsten code generation with optimized paths:

DecimalType with scale 0: Generates identity code ($c)
DecimalType with scale > 0: Generates $c.floor() method call
LongType: Generates identity code ($c)
Other types: Generates (long)(java.lang.Math.FLOOR($c)) with explicit casting

Examples¶

-- Basic floor operations
SELECT FLOOR(3.7);        -- Returns 3
SELECT FLOOR(-2.1);       -- Returns -3
SELECT FLOOR(5.0);        -- Returns 5
SELECT FLOOR(42);         -- Returns 42

-- With decimal types
SELECT FLOOR(CAST(123.456 AS DECIMAL(6,3)));  -- Returns 123 as DECIMAL(4,0)

// DataFrame API examples
import org.apache.spark.sql.functions.floor

// Apply floor to a numeric column
df.select(floor(col("price")))

// Floor operation in transformations
df.withColumn("price_floor", floor(col("price")))

// Combined with other operations
df.select(floor(col("rating") * 2.5))