Uniform¶

Overview¶

The uniform function generates random floating-point numbers from a uniform distribution within a specified range. It returns pseudo-random values between the minimum and maximum bounds (inclusive of min, exclusive of max) using an optional seed for reproducible results.

Syntax¶

uniform(min, max)
uniform(min, max, seed)

Arguments¶

Argument	Type	Description
min	Numeric	The minimum bound of the uniform distribution (inclusive)
max	Numeric	The maximum bound of the uniform distribution (exclusive)
seed	Integer/Long	Optional seed value for reproducible random number generation

Return Type¶

The return type follows numeric type precedence rules:

For integral types: returns the wider of the two input types
For floating-point types: returns DoubleType when either input is DoubleType, otherwise FloatType
For decimal types: returns the wider decimal type
Mixed numeric types follow standard SQL numeric precedence

Supported Data Types¶

Input types: All numeric types (IntegerType, LongType, FloatType, DoubleType, DecimalType)
Seed types: IntegerType or LongType only
Output: Numeric type based on input type precedence

Algorithm¶

Validates that all inputs are foldable (constant expressions)
Uses the internal Rand function with the provided or default seed
Implements the formula: min + (max - min) * rand(seed)
Casts inputs to DoubleType for calculation, then casts result back to target type
Returns null if any input expression evaluates to null

Partitioning Behavior¶

This expression does not affect partitioning behavior:

Preserves existing partitioning schemes
Does not require data shuffle
Generates independent random values per partition

Edge Cases¶

Null handling: Returns null if any input parameter (min, max, seed) is null
Non-foldable inputs: Throws DataTypeMismatch error if min, max, or seed are not constant expressions
Type resolution: Returns NullType if seed expression is unresolved or NullType
Invalid ranges: No explicit validation for min >= max cases in the expression itself
Deterministic behavior: Non-deterministic by nature, but reproducible with same seed

Code Generation¶

This expression uses runtime replacement and does not directly support Tungsten code generation. It is replaced at runtime with a combination of Add, Multiply, Subtract, Cast, and Rand expressions that may support code generation.

Examples¶

-- Generate random number between 10 and 20
SELECT uniform(10, 20) AS random_value;

-- Generate reproducible random number with seed
SELECT uniform(0.0, 1.0, 12345) AS seeded_random;

-- Use with different numeric types
SELECT uniform(1, 100, 0) AS int_range;
SELECT uniform(1.5, 2.5, 42) AS float_range;

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

// Generate random values in range
df.select(expr("uniform(1, 10)").alias("random_int"))

// With seed for reproducibility  
df.select(expr("uniform(0.0, 1.0, 123)").alias("seeded_random"))