ArrayMax¶

Overview¶

The ArrayMax expression returns the maximum element from an array. It compares all non-null elements in the input array using the natural ordering of the element data type and returns the largest value found.

Syntax¶

array_max(array_expr)

// DataFrame API
col("array_column").expr("array_max(array_column)")

Arguments¶

Argument	Type	Description
array_expr	ArrayType	Input array from which to find the maximum element

Return Type¶

Returns the same data type as the array elements. For example, if the input is array<int>, the return type is int. If the input is array<string>, the return type is string.

Supported Data Types¶

Supports arrays containing any orderable data types:

Numeric types (byte, short, int, long, float, double, decimal)
String types (string, varchar, char)
Date and timestamp types
Boolean type
Binary type

The element type must support ordering comparison operations.

Algorithm¶

The expression evaluates the maximum using the following approach:

Iterates through each element in the input array
Skips null elements during comparison
Maintains the current maximum value using the data type's natural ordering
Uses TypeUtils.getInterpretedOrdering() to perform element comparisons
Returns the final maximum value or null if no valid elements exist

Partitioning Behavior¶

This expression does not affect partitioning behavior:

Preserves existing partitioning schemes
Does not require data shuffle as it operates on individual array values
Can be pushed down to individual partitions for parallel execution

Edge Cases¶

Null array input: Returns null if the entire input array is null
Empty array: Returns null for empty arrays
All null elements: Returns null if all array elements are null
Mixed null/non-null: Ignores null elements and finds maximum among non-null values
Single element: Returns that element (or null if the single element is null)
Duplicate maximums: Returns one instance of the maximum value

Code Generation¶

This expression supports Tungsten code generation for optimized performance. The generated code:

Uses efficient loop iteration over array elements
Leverages ctx.reassignIfGreater() for optimized comparison logic
Avoids boxing/unboxing overhead for primitive types
Falls back to interpreted evaluation (nullSafeEval) when code generation is disabled

Examples¶

-- Basic numeric array
SELECT array_max(array(1, 20, null, 3));
-- Result: 20

-- String array  
SELECT array_max(array('apple', 'zebra', 'banana'));
-- Result: 'zebra'

-- Array with all nulls
SELECT array_max(array(null, null));
-- Result: null

-- Empty array
SELECT array_max(array());
-- Result: null

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

df.select(expr("array_max(numbers_array)").as("max_value"))

// With array literal
df.select(expr("array_max(array(1, 5, 3, 9, 2))").as("max_num"))