[SPARK-57736][SQL] Fix NPE in CreateNamedStruct.dataType when a struct field name is null

[MaxGekk]

What changes were proposed in this pull request?

CreateNamedStruct.dataType builds each field with StructField(name.toString, ...):

override lazy val dataType: StructType = {
  val fields = names.zip(valExprs).map {
    case (name, expr) =>
      ...
      StructField(name.toString, expr.dataType, expr.nullable, metadata)   // NPE if name == null
  }
  StructType(fields)
}

When a field name is null, name.toString throws a NullPointerException. This is reached eagerly while building a RowEncoder serializer (SerializerBuildHelper.createSerializerForObject -> CreateNamedStruct(...).dataType), so it crashes before any analysis runs. This PR makes the field name null-safe and preserves the null name:

StructField(if (name == null) null else name.toString, expr.dataType, expr.nullable, metadata)

Why are the changes needed?

A null field name is invalid input -- CreateNamedStruct.checkInputDataTypes already rejects it (names.contains(null) -> UNEXPECTED_NULL) -- but dataType dereferences name.toString before type checking, and the encoder calls dataType directly. Keeping it null-safe converts the hard NullPointerException into correct behavior, consistent with SPARK-57725 which made AttributeSeq tolerate null-named attributes.

Minimal reproduction:

import org.apache.spark.sql.catalyst.expressions.{CreateNamedStruct, Literal}
import org.apache.spark.sql.types.{IntegerType, StringType}

CreateNamedStruct(Seq(Literal.create(null, StringType), Literal(1))).dataType  // NPE before this fix

Note: this fixes the specific CreateNamedStruct.dataType NPE. The full createDataFrame(schemaWithNullFieldName) scenario hits additional, independent null-name sites further along (e.g. a StructField.name.equalsIgnoreCase schema comparison during resolution), which are separate pre-existing issues and out of scope here.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

Added a regression test in ComplexTypeSuite asserting dataType no longer throws and preserves the null field name.

build/sbt 'catalyst/testOnly *ComplexTypeSuite'

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

Generated-by: Cursor

[MaxGekk]

Could you review when you have a moment? cc @cloud-fan @dongjoon-hyun

[uros-b]

Please update the test a bit - the current regression test asserts dt.length == 1 and dt.head.name == null (core behavior covered) but does not assert checkInputDataTypes() still returns UNEXPECTED_NULL, nor exercise a null name mixed with valid named fields.

[uros-b]

Thank you @MaxGekk, I left one note but otherwise looks good!

[MaxGekk]

Thank you for the review, @uros-b! Addressed your comment: the regression test now also asserts checkInputDataTypes() returns UNEXPECTED_NULL and exercises a null field name mixed with valid named fields.

[dongjoon-hyun]

+1, LGTM.

[MaxGekk]

Merging to master/4.x/4.2/4.1/4.0. Thank you, @uros-b and @dongjoon-hyun for review.