If¶

Overview¶

The If expression is a conditional expression that evaluates a boolean predicate and returns one of two values based on the result. It implements the standard ternary conditional logic where a true predicate returns the first value, and a false predicate returns the second value.

Syntax¶

IF(condition, true_value, false_value)

Arguments¶

Argument	Type	Description
predicate	Boolean	The condition to evaluate
trueValue	Any	The value to return if the predicate is true
falseValue	Any	The value to return if the predicate is false

Return Type¶

Returns the unified data type of the trueValue and falseValue expressions after type merging. The return type is nullable if either the true or false value expressions are nullable.

Supported Data Types¶

Predicate must be of BooleanType
True and false value expressions must have compatible types that can be merged through type coercion
All primitive and complex data types are supported for the value expressions

Algorithm¶

The predicate expression is always evaluated first
If the predicate evaluates to true, the trueValue expression is evaluated and returned
If the predicate evaluates to false or null, the falseValue expression is evaluated and returned
Only one of the value expressions is evaluated per row (lazy evaluation)
Type coercion is applied to ensure both value expressions have the same data type

Partitioning Behavior¶

The If expression preserves partitioning characteristics:

Does not require shuffle operations
Maintains existing partitioning schemes
Does not affect data distribution across partitions

Edge Cases¶

If the predicate is null, the expression returns the false value
The result is null only if the selected branch (true or false value) evaluates to null
Type validation ensures both value expressions can be coerced to a common type
The predicate must be exactly BooleanType - no implicit conversion from other types

Code Generation¶

This expression supports full Tungsten code generation through the doGenCode method. It generates efficient Java code that:

Evaluates the predicate condition first
Uses conditional branching to evaluate only the selected value expression
Properly handles null values and type conversions

Examples¶

-- Basic conditional logic
SELECT IF(age >= 18, 'Adult', 'Minor') AS category FROM users;

-- Null handling
SELECT IF(score IS NOT NULL, score, 0) AS final_score FROM tests;

-- Nested conditions
SELECT IF(status = 'ACTIVE', IF(premium, 'Premium User', 'Regular User'), 'Inactive') FROM accounts;

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

df.select(when(col("age") >= 18, "Adult").otherwise("Minor").as("category"))

// Using expr for IF function
df.select(expr("IF(score IS NOT NULL, score, 0)").as("final_score"))

If¶