[SPARK-48871] Fix INVALID_NON_DETERMINISTIC_EXPRESSIONS validation in…

… CheckAnalysis

### What changes were proposed in this pull request?

The PR added a trait that logical plans can extend to implement a method to decide whether there can be non-deterministic expressions for the operator, and check this method in checkAnalysis.

### Why are the changes needed?

I encountered the `INVALID_NON_DETERMINISTIC_EXPRESSIONS` exception when attempting to use a non-deterministic udf in my query. The non-deterministic expression can be safely allowed for my custom LogicalPlan, but it is disabled in the checkAnalysis phase. The CheckAnalysis rule is too strict so that reasonable use cases of non-deterministic expressions are also disabled.

### Does this PR introduce _any_ user-facing change?

No

### How was this patch tested?

The test case `"SPARK-48871: AllowsNonDeterministicExpression allow lists non-deterministic expressions"` is added.

### Was this patch authored or co-authored using generative AI tooling?

No

Closes apache#47304 from zhipengmao-db/zhipengmao-db/SPARK-48871-check-analysis.

Lead-authored-by: zhipeng.mao <zhipeng.mao@databricks.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

Author: 2 people

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/CheckAnalysis.scala

@@ -143,6 +143,17 @@ trait CheckAnalysis extends PredicateHelper with LookupCatalog with QueryErrorsB
       errorClass, missingCol, orderedCandidates, a.origin)
   }
 
+  /**
+   * Checks whether the operator allows non-deterministic expressions.
+   */
+  private def operatorAllowsNonDeterministicExpressions(plan: LogicalPlan): Boolean = {
+    plan match {
+      case p: SupportsNonDeterministicExpression =>
+        p.allowNonDeterministicExpression
+      case _ => false
+    }
+  }
+
   def checkAnalysis(plan: LogicalPlan): Unit = {
     // We should inline all CTE relations to restore the original plan shape, as the analysis check
     // may need to match certain plan shapes. For dangling CTE relations, they will still be kept
@@ -718,6 +729,7 @@ trait CheckAnalysis extends PredicateHelper with LookupCatalog with QueryErrorsB
                 "dataType" -> toSQLType(mapCol.dataType)))
 
           case o if o.expressions.exists(!_.deterministic) &&
+            !operatorAllowsNonDeterministicExpressions(o) &&
             !o.isInstanceOf[Project] &&
             // non-deterministic expressions inside CollectMetrics have been
             // already validated inside checkMetric function

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicLogicalOperators.scala

@@ -1998,6 +1998,16 @@ case class DeduplicateWithinWatermark(keys: Seq[Attribute], child: LogicalPlan)
  */
 trait SupportsSubquery extends LogicalPlan
 
+/**
+ * Trait that logical plans can extend to check whether it can allow non-deterministic
+ * expressions and pass the CheckAnalysis rule.
+ */
+trait SupportsNonDeterministicExpression extends LogicalPlan {
+
+  /** Returns whether it allows non-deterministic expressions. */
+  def allowNonDeterministicExpression: Boolean
+}
+
 /**
  * Collect arbitrary (named) metrics from a dataset. As soon as the query reaches a completion
  * point (batch query completes or streaming query epoch completes) an event is emitted on the

sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/AnalysisErrorSuite.scala

@@ -90,6 +90,13 @@ case class TestFunctionWithTypeCheckFailure(
 
 case class UnresolvedTestPlan() extends UnresolvedLeafNode
 
+case class SupportsNonDeterministicExpressionTestOperator(
+    actions: Seq[Expression],
+    allowNonDeterministicExpression: Boolean)
+  extends LeafNode with SupportsNonDeterministicExpression {
+  override def output: Seq[Attribute] = Seq()
+}
+
 class AnalysisErrorSuite extends AnalysisTest with DataTypeErrorsBase {
   import TestRelations._
 
@@ -1364,4 +1371,20 @@ class AnalysisErrorSuite extends AnalysisTest with DataTypeErrorsBase {
     messageParameters = Map(
       "expr" -> "\"_w0\"",
       "exprType" -> "\"MAP<STRING, STRING>\""))
+
+  test("SPARK-48871: SupportsNonDeterministicExpression allows non-deterministic expressions") {
+    val nonDeterministicExpressions = Seq(new Rand())
+    val tolerantPlan =
+      SupportsNonDeterministicExpressionTestOperator(
+        nonDeterministicExpressions, allowNonDeterministicExpression = true)
+    assertAnalysisSuccess(tolerantPlan)
+
+    val intolerantPlan =
+      SupportsNonDeterministicExpressionTestOperator(
+        nonDeterministicExpressions, allowNonDeterministicExpression = false)
+    assertAnalysisError(
+      intolerantPlan,
+      "INVALID_NON_DETERMINISTIC_EXPRESSIONS" :: Nil
+    )
+  }
 }