DivideDTInterval¶

Overview¶

The DivideDTInterval expression divides a day-time interval by a numeric value, returning a new day-time interval. This operation performs division with HALF_UP rounding mode and includes comprehensive overflow and divide-by-zero checks to ensure safe arithmetic operations.

Syntax¶

day_time_interval / numeric_value

Arguments¶

Argument	Type	Description
`interval`	`DayTimeIntervalType`	The day-time interval to be divided
`num`	`NumericType`	The numeric divisor (integral, decimal, or fractional types)

Return Type¶

DayTimeIntervalType() - Returns a day-time interval representing the result of the division operation.

Supported Data Types¶

Left operand (interval): DayTimeIntervalType only
Right operand (num): All numeric types including:
IntegralType (Byte, Short, Int, Long)
DecimalType
FractionalType (Float, Double)

Algorithm¶

Converts the day-time interval to microseconds (Long representation)
Performs type-specific division based on the divisor's data type:
Integral types: Uses LongMath.divide() with HALF_UP rounding
Decimal types: Converts to Decimal, performs division, then rounds to Long with HALF_UP
Fractional types: Uses DoubleMath.roundToLong() with HALF_UP rounding
Validates for divide-by-zero and overflow conditions before computation
Returns the result as a day-time interval in microseconds

Partitioning Behavior¶

Preserves partitioning: No, this is a value-transforming operation
Requires shuffle: No, operates on individual rows independently
Can be applied within existing partitions without data movement

Edge Cases¶

Null handling: Expression is null-intolerant (nullIntolerant = true) - returns null if either operand is null
Divide by zero: Throws QueryExecutionErrors when divisor is zero
Overflow behavior:
Checks for Long.MinValue / -1 overflow condition
Throws QueryExecutionErrors.overflowInIntegralDivideError on overflow
Uses longValueExact() for decimal operations to detect overflow
Precision: All division operations use HALF_UP rounding mode for consistent behavior

Code Generation¶

This expression supports full Tungsten code generation with type-specific optimized paths: - Integral types: Generates inline overflow checks and LongMath.divide() calls - Decimal types: Generates Decimal arithmetic with precision handling - Fractional types: Generates optimized double division with DoubleMath.roundToLong() - Includes generated divide-by-zero validation for all numeric types

Examples¶

-- Divide a day-time interval by an integer
SELECT INTERVAL '2 12:30:45.123' DAY TO SECOND / 2;
-- Result: INTERVAL '1 06:15:22.561500' DAY TO SECOND

-- Divide by decimal value
SELECT INTERVAL '5 00:00:00' DAY TO SECOND / 2.5;
-- Result: INTERVAL '2 00:00:00' DAY TO SECOND

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

// Divide interval column by numeric literal
df.select(col("day_time_interval") / lit(3))

// Divide interval by another numeric column
df.select(col("day_time_interval") / col("divisor"))