Log¶

Overview¶

The Log expression computes the natural logarithm (base e) of a numeric value using StrictMath.log. It extends UnaryLogExpression and provides the mathematical logarithm function with alias support for "ln".

Syntax¶

LOG(expr)
LN(expr)  -- alias

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

Arguments¶

Argument	Type	Description
child	Expression	The numeric expression to compute the natural logarithm of

Return Type¶

Double - Returns a double precision floating point number representing the natural logarithm.

Supported Data Types¶

Numeric types (Int, Long, Float, Double, Decimal)
Values are internally converted to double for computation

Algorithm¶

Evaluates the child expression to get the input value
Converts the input to double precision
Applies StrictMath.log() which computes the natural logarithm (base e)
Returns the result as a double value
Handles special mathematical cases according to StrictMath.log behavior

Partitioning Behavior¶

Preserves partitioning as it's a deterministic unary expression
Does not require shuffle operations
Can be pushed down to individual partitions independently

Edge Cases¶

Returns NaN for negative input values
Returns Double.NEGATIVE_INFINITY for input value 0.0
Returns NaN for NaN input
Returns Double.POSITIVE_INFINITY for Double.POSITIVE_INFINITY input
Null input values result in null output

Code Generation¶

This expression supports Catalyst code generation (Tungsten) through its parent UnaryLogExpression class, generating efficient Java bytecode for evaluation.

Examples¶

-- Basic natural logarithm
SELECT LOG(2.718281828);
-- Result: ~1.0

SELECT LN(1);
-- Result: 0.0

-- Edge cases
SELECT LOG(-1);
-- Result: NaN

SELECT LOG(0);
-- Result: -Infinity

// DataFrame API usage
import org.apache.spark.sql.functions._

val df = spark.range(1, 10).toDF("value")
df.select(log(col("value"))).show()

// Using alias
df.select(expr("ln(value)")).show()