TryEval¶

Overview¶

TryEval is a Spark Catalyst unary expression that wraps another expression in exception handling logic. It evaluates its child expression and returns the result if successful, or null if any exception occurs during evaluation. This expression provides a safe way to handle potentially failing operations without causing query failures.

Syntax¶

This is an internal Catalyst expression primarily used within Spark's query engine. It may be generated automatically by the optimizer or used in DataFrame operations that require exception-safe evaluation.

// Internal usage in Catalyst expressions
TryEval(childExpression)

Arguments¶

Argument	Type	Description
child	Expression	The child expression to evaluate within a try-catch block

Return Type¶

Returns the same data type as the child expression, but the result is always nullable regardless of the child's nullability.

Supported Data Types¶

Supports all data types that the child expression supports, including: - All primitive types (Boolean, Byte, Short, Integer, Long, Float, Double) - Complex types (String, Binary, Array, Map, Struct) - Date/Time types (Date, Timestamp, CalendarInterval) - Decimal types

Algorithm¶

Initialize result variables with null flag set to true and default value for the data type
Execute the child expression within a try-catch block
If evaluation succeeds: copy the child's null flag and result value
If any exception occurs: leave result as null (isNull = true)
Return the final result with appropriate null flag

Partitioning Behavior¶

Preserves partitioning: Yes, this expression does not change data distribution
Requires shuffle: No, evaluation is performed locally on each partition
Acts as a pass-through for partitioning schemes since it only affects individual row values

Edge Cases¶

Null handling: If child expression returns null normally, TryEval preserves that null
Exception handling: Any exception (RuntimeException, ArithmeticException, etc.) results in null output
Nested exceptions: Catches all Exception types, including those from deeply nested expression trees
Nullability override: Always returns nullable results even if child expression is non-nullable
Code generation: Exception handling is preserved in generated code with proper try-catch blocks

Code Generation¶

This expression supports full code generation (Tungsten). The generated code includes: - Proper Java try-catch blocks around child expression evaluation - Initialization of null flags and default values - Exception-safe variable assignments - No fallback to interpreted mode required

Examples¶

-- TryEval is typically generated internally, not directly accessible in SQL
-- However, it might be used internally for operations like:
SELECT CASE WHEN some_condition THEN risky_operation(col) END FROM table

// Internal Catalyst usage example
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.types._

// Wrap a potentially failing expression
val riskyExpr = Divide(Literal(10), col("denominator"))
val safeExpr = TryEval(riskyExpr)

// The safe expression will return null instead of throwing division by zero