Update the java code to properly deal with lists being returned as st…

…rings (#16536)

Recently some JSON parsing was updated so lists could be returned as strings. This updates the java code so that when cleaning up the results to match the desired schema that it can handle corner cases associated with lists and structs properly.

Tests are covered in the Spark plugin, but I am happy to add some here if we really want to validate that part of this.

Authors:
  - Robert (Bobby) Evans (https://github.com/revans2)

Approvers:
  - Nghia Truong (https://github.com/ttnghia)

URL: #16536

java/src/main/java/ai/rapids/cudf/Table.java

@@ -1084,7 +1084,12 @@ private static DidViewChange gatherJSONColumns(Schema schema, TableWithMeta.Nest
         // The types don't match so just return the input unchanged...
         return DidViewChange.no();
       } else {
-        String[] foundNames = children.getNames();
+        String[] foundNames;
+        if (children == null) {
+          foundNames = new String[0];
+        } else {
+          foundNames = children.getNames();
+        }
         HashMap<String, Integer> indices = new HashMap<>();
         for (int i = 0; i < foundNames.length; i++) {
           indices.put(foundNames[i], i);
@@ -1101,8 +1106,9 @@ private static DidViewChange gatherJSONColumns(Schema schema, TableWithMeta.Nest
           for (int i = 0; i < columns.length; i++) {
             String neededColumnName = neededNames[i];
             Integer index = indices.get(neededColumnName);
+            Schema childSchema = schema.getChild(i);
             if (index != null) {
-              if (schema.getChild(i).isStructOrHasStructDescendant()) {
+              if (childSchema.isStructOrHasStructDescendant()) {
                 ColumnView child = cv.getChildColumnView(index);
                 boolean shouldCloseChild = true;
                 try {
@@ -1131,8 +1137,23 @@ private static DidViewChange gatherJSONColumns(Schema schema, TableWithMeta.Nest
               }
             } else {
               somethingChanged = true;
-              try (Scalar s = Scalar.fromNull(types[i])) {
-                columns[i] = ColumnVector.fromScalar(s, (int) cv.getRowCount());
+              if (types[i] == DType.LIST) {
+                try (Scalar s = Scalar.listFromNull(childSchema.getChild(0).asHostDataType())) {
+                  columns[i] = ColumnVector.fromScalar(s, (int) cv.getRowCount());
+                }
+              } else if (types[i] == DType.STRUCT) {
+                int numStructChildren = childSchema.getNumChildren();
+                HostColumnVector.DataType[] structChildren = new HostColumnVector.DataType[numStructChildren];
+                for (int structChildIndex = 0; structChildIndex < numStructChildren; structChildIndex++) {
+                  structChildren[structChildIndex] = childSchema.getChild(structChildIndex).asHostDataType();
+                }
+                try (Scalar s = Scalar.structFromNull(structChildren)) {
+                  columns[i] = ColumnVector.fromScalar(s, (int) cv.getRowCount());
+                }
+              } else {
+                try (Scalar s = Scalar.fromNull(types[i])) {
+                  columns[i] = ColumnVector.fromScalar(s, (int) cv.getRowCount());
+                }
               }
             }
           }