[SPARK-55696][SQL] Add explicit error to Encoders.bean for interface class

[szehon-ho]

What changes were proposed in this pull request?

Add explicit error message to Encoders.bean against interfaces

Why are the changes needed?

For bean Encoders, the de-serializer uses the constructor inferred from the bean Encoder. ie, we get the bytes back, the Dataset deserialization will use the Encoder's interface class to try to construct , and fail.

Code ref:

spark/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/DeserializerBuildHelper.scala

Line 482 in b370375

val newInstance = NewInstance(cls, Nil, ObjectType(cls), propagateNull = false)

Does this PR introduce any user-facing change?

No, it already fails today with another message:

java.lang.IllegalStateException: found an unhandled type: null
      at org.apache.commons.lang3.reflect.TypeUtils.getTypeArguments(TypeUtils.java:915)
      at org.apache.commons.lang3.reflect.TypeUtils.getTypeArguments(TypeUtils.java:791)
      at org.apache.commons.lang3.reflect.TypeUtils.getTypeArguments(TypeUtils.java:886)
      at org.apache.commons.lang3.reflect.TypeUtils.getTypeArguments(TypeUtils.java:873)
      at org.apache.spark.sql.catalyst.JavaTypeInference$.encoderFor(JavaTypeInference.scala:159)
      at org.apache.spark.sql.catalyst.JavaTypeInference$.encoderFor(JavaTypeInference.scala:63)
      at org.apache.spark.sql.catalyst.JavaTypeInference$.encoderFor(JavaTypeInference.scala:56)
      at org.apache.spark.sql.Encoders$.bean(Encoders.scala:211)
      at org.apache.spark.sql.Encoders.bean(Encoders.scala)

it will now fail more explicitly.

How was this patch tested?

Added unit test in JavaDatasetSuite

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

Yes, cursor

[cloud-fan]

if it's only used once we can inline it.

[szehon-ho]

done

[pan3793]

Hmm... why add a LEGACY error condition?