fix(spark): casting date/time requires timezone (#318)

When casting a date type to a string, Spark requires that a timezone is specified,
otherwise it will not resolve the logical plan.
The timezone is ignored for non-date/time values.

Signed-off-by: Andrew Coleman <andrew_coleman@uk.ibm.com>

Author: andrew-coleman

spark/src/main/scala/io/substrait/spark/expression/ToSparkExpression.scala

@@ -20,7 +20,8 @@ import io.substrait.spark.{DefaultExpressionVisitor, HasOutputStack, SparkExtens
 import io.substrait.spark.logical.ToLogicalPlan
 
 import org.apache.spark.sql.catalyst.expressions.{CaseWhen, Cast, Expression, In, Literal, MakeDecimal, NamedExpression, ScalarSubquery}
-import org.apache.spark.sql.types.Decimal
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.sql.types.{DateType, Decimal}
 import org.apache.spark.substrait.SparkTypeUtil
 import org.apache.spark.unsafe.types.UTF8String
 
@@ -153,7 +154,12 @@ class ToSparkExpression(
 
   override def visit(expr: SExpression.Cast): Expression = {
     val childExp = expr.input().accept(this)
-    Cast(childExp, ToSubstraitType.convert(expr.getType))
+    val tt = ToSubstraitType.convert(expr.getType)
+    val tz = childExp.dataType match {
+      case DateType => Some(SQLConf.get.getConf(SQLConf.SESSION_LOCAL_TIMEZONE))
+      case _ => None
+    }
+    Cast(childExp, tt, tz)
   }
 
   override def visit(expr: exp.FieldReference): Expression = {
@@ -197,6 +203,7 @@ class ToSparkExpression(
     val list = expr.options().asScala.map(e => e.accept(this))
     In(value, list)
   }
+
   override def visit(expr: SExpression.ScalarFunctionInvocation): Expression = {
     val eArgs = expr.arguments().asScala
     val args = eArgs.zipWithIndex.map {

spark/src/test/scala/io/substrait/spark/TPCDSPlan.scala

@@ -36,7 +36,7 @@ class TPCDSPlan extends TPCDSBase with SubstraitPlanTestBase {
     "q2", // because round() isn't defined in substrait to work with Decimal. https://github.com/substrait-io/substrait/pull/713
     "q9", // requires implementation of named_struct()
     "q10", "q35", "q45", // Unsupported join type ExistenceJoin (this is an internal spark type)
-    "q51", "q83", "q84", // TBD
+    "q51", "q84", // TBD
     "q72" //requires implementation of date_add()
   )
   // spotless:on