diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactory.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactory.java
index e003ee7..bff80bc 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactory.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactory.java
@@ -18,7 +18,7 @@
 import jp.co.yahoo.yosegi.inmemory.ILoader;
 import jp.co.yahoo.yosegi.inmemory.ILoaderFactory;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkArrayLoader;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyArrayLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkRunLengthEncodingArrayLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkUnionArrayLoader;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
@@ -37,7 +37,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
       throws IOException {
     if (columnBinary == null) {
       // FIXME:
-      return new SparkNullLoader(vector, loadSize);
+      return new SparkEmptyArrayLoader(vector, loadSize);
     }
     switch (getLoadType(columnBinary, loadSize)) {
       case ARRAY:
@@ -47,8 +47,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
       case UNION:
         return new SparkUnionArrayLoader(vector, loadSize);
       default:
-        // FIXME:
-        return new SparkNullLoader(vector, loadSize);
+        return new SparkEmptyArrayLoader(vector, loadSize);
     }
   }
 }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactory.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactory.java
index b6a38ba..fc06b4c 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactory.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactory.java
@@ -17,8 +17,8 @@
 import jp.co.yahoo.yosegi.binary.ColumnBinary;
 import jp.co.yahoo.yosegi.inmemory.ILoader;
 import jp.co.yahoo.yosegi.inmemory.ILoaderFactory;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyMapLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkMapLoader;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkUnionMapLoader;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
 
@@ -36,7 +36,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
       throws IOException {
     if (columnBinary == null) {
       // FIXME:
-      return new SparkNullLoader(vector, loadSize);
+      return new SparkEmptyMapLoader(vector, loadSize);
     }
     switch (getLoadType(columnBinary, loadSize)) {
       case SPREAD:
@@ -45,7 +45,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
         return new SparkUnionMapLoader(vector, loadSize);
       default:
         // FIXME:
-        return new SparkNullLoader(vector, loadSize);
+        return new SparkEmptyMapLoader(vector, loadSize);
     }
   }
 }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactory.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactory.java
index 3c4fb3d..0f0b9f8 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactory.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactory.java
@@ -17,7 +17,7 @@
 import jp.co.yahoo.yosegi.binary.ColumnBinary;
 import jp.co.yahoo.yosegi.inmemory.ILoader;
 import jp.co.yahoo.yosegi.inmemory.ILoaderFactory;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyStructLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkStructLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkUnionStructLoader;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
@@ -36,7 +36,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
       throws IOException {
     if (columnBinary == null) {
       // FIXME:
-      return new SparkNullLoader(vector, loadSize);
+      return new SparkEmptyStructLoader(vector, loadSize);
     }
     switch (getLoadType(columnBinary, loadSize)) {
       case SPREAD:
@@ -45,7 +45,7 @@ public ILoader createLoader(final ColumnBinary columnBinary, final int loadSize)
         return new SparkUnionStructLoader(vector, loadSize);
       default:
         // FIXME:
-        return new SparkNullLoader(vector, loadSize);
+        return new SparkEmptyStructLoader(vector, loadSize);
     }
   }
 }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkArrayLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkArrayLoader.java
index f99512a..a272314 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkArrayLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkArrayLoader.java
@@ -38,7 +38,7 @@ public int getLoadSize() {
 
   @Override
   public void setNull(final int index) throws IOException {
-    vector.putNull(index);
+    vector.putArray(index, 0, 0);
   }
 
   @Override
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoader.java
new file mode 100644
index 0000000..29ad378
--- /dev/null
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoader.java
@@ -0,0 +1,71 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one or more contributor license
+ * agreements. See the NOTICE file distributed with this work for additional information regarding
+ * copyright ownership. The ASF licenses this file to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance with the License. You may obtain a
+ * copy of the License at
+ *
+ * <p>http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * <p>Unless required by applicable law or agreed to in writing, software distributed under the
+ * License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
+ * express or implied. See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import jp.co.yahoo.yosegi.inmemory.LoadType;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+
+import java.io.IOException;
+
+public class SparkEmptyArrayLoader implements ILoader<WritableColumnVector> {
+
+  private final WritableColumnVector vector;
+  private final int loadSize;
+
+  public SparkEmptyArrayLoader(final WritableColumnVector vector, final int loadSize) {
+    this.vector = vector;
+    this.loadSize = loadSize;
+    this.vector.getChild(0).reset();
+    this.vector.getChild(0).reserve(0);
+    if (this.vector.getChild(0).hasDictionary()) {
+      this.vector.getChild(0).reserveDictionaryIds(0);
+      this.vector.getChild(0).setDictionary(null);
+    }
+  }
+
+  @Override
+  public LoadType getLoaderType() {
+    return LoadType.NULL;
+  }
+
+  @Override
+  public int getLoadSize() {
+    return loadSize;
+  }
+
+  @Override
+  public void setNull(final int index) throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public void finish() throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public WritableColumnVector build() throws IOException {
+    for (int i = 0; i < loadSize; i++) {
+      vector.putArray(i, 0, 0);
+    }
+    return vector;
+  }
+
+  @Override
+  public boolean isLoadingSkipped() {
+    return true;
+  }
+}
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoader.java
new file mode 100644
index 0000000..7669968
--- /dev/null
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoader.java
@@ -0,0 +1,23 @@
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.ArrayType;
+import org.apache.spark.sql.types.MapType;
+import org.apache.spark.sql.types.StructType;
+
+import java.io.IOException;
+
+public class SparkEmptyLoader {
+    public static void load(final WritableColumnVector vector, final int loadSize) throws IOException {
+        final Class klass = vector.dataType().getClass();
+        if (klass == ArrayType.class) {
+            new SparkEmptyArrayLoader(vector, loadSize).build();
+        } else if (klass == StructType.class) {
+            new SparkEmptyStructLoader(vector, loadSize).build();
+        } else if (klass == MapType.class) {
+            new SparkEmptyMapLoader(vector, loadSize).build();
+        } else {
+            new SparkNullLoader(vector, loadSize).build();
+        }
+    }
+}
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoader.java
new file mode 100644
index 0000000..7e2f598
--- /dev/null
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoader.java
@@ -0,0 +1,66 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one or more contributor license
+ * agreements. See the NOTICE file distributed with this work for additional information regarding
+ * copyright ownership. The ASF licenses this file to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance with the License. You may obtain a
+ * copy of the License at
+ *
+ * <p>http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * <p>Unless required by applicable law or agreed to in writing, software distributed under the
+ * License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
+ * express or implied. See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import jp.co.yahoo.yosegi.inmemory.LoadType;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+
+import java.io.IOException;
+
+public class SparkEmptyMapLoader implements ILoader<WritableColumnVector> {
+
+  private final WritableColumnVector vector;
+  private final int loadSize;
+
+  public SparkEmptyMapLoader(final WritableColumnVector vector, final int loadSize) {
+    this.vector = vector;
+    this.loadSize = loadSize;
+  }
+
+  @Override
+  public LoadType getLoaderType() {
+    return LoadType.NULL;
+  }
+
+  @Override
+  public int getLoadSize() {
+    return loadSize;
+  }
+
+  @Override
+  public void setNull(final int index) throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public void finish() throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public WritableColumnVector build() throws IOException {
+    vector.getChild(0).reset();
+    vector.getChild(0).reserve(0);
+    vector.getChild(1).reset();
+    vector.getChild(1).reserve(0);
+    return vector;
+  }
+
+  @Override
+  public boolean isLoadingSkipped() {
+    return true;
+  }
+}
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoader.java
new file mode 100644
index 0000000..cabc740
--- /dev/null
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoader.java
@@ -0,0 +1,80 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one or more contributor license
+ * agreements. See the NOTICE file distributed with this work for additional information regarding
+ * copyright ownership. The ASF licenses this file to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance with the License. You may obtain a
+ * copy of the License at
+ *
+ * <p>http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * <p>Unless required by applicable law or agreed to in writing, software distributed under the
+ * License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
+ * express or implied. See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import jp.co.yahoo.yosegi.inmemory.LoadType;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.ArrayType;
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.MapType;
+import org.apache.spark.sql.types.StructType;
+
+import java.io.IOException;
+
+public class SparkEmptyStructLoader implements ILoader<WritableColumnVector> {
+
+  private final WritableColumnVector vector;
+  private final int loadSize;
+  private final String[] names;
+
+  public SparkEmptyStructLoader(final WritableColumnVector vector, final int loadSize) {
+    this.vector = vector;
+    this.loadSize = loadSize;
+    final StructType structType = (StructType) vector.dataType();
+    this.names = structType.fieldNames();
+    for (int i = 0; i < names.length; i++) {
+      vector.getChild(i).reset();
+      vector.getChild(i).reserve(loadSize);
+      if (vector.getChild(i).hasDictionary()) {
+        vector.getChild(i).reserveDictionaryIds(0);
+        vector.getChild(i).setDictionary(null);
+      }
+    }
+  }
+
+  @Override
+  public LoadType getLoaderType() {
+    return LoadType.NULL;
+  }
+
+  @Override
+  public int getLoadSize() {
+    return loadSize;
+  }
+
+  @Override
+  public void setNull(final int index) throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public void finish() throws IOException {
+    // FIXME:
+  }
+
+  @Override
+  public WritableColumnVector build() throws IOException {
+    for (int i = 0; i < names.length; i++) {
+      SparkEmptyLoader.load(vector.getChild(i), loadSize);
+    }
+    return vector;
+  }
+
+  @Override
+  public boolean isLoadingSkipped() {
+    return true;
+  }
+}
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoader.java
index d8d04d1..9f8112f 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoader.java
@@ -14,13 +14,13 @@
  */
 package jp.co.yahoo.yosegi.spark.inmemory.loader;
 
-import jp.co.yahoo.yosegi.inmemory.ISequentialLoader;
+import jp.co.yahoo.yosegi.inmemory.ILoader;
 import jp.co.yahoo.yosegi.inmemory.LoadType;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
 
 import java.io.IOException;
 
-public class SparkNullLoader implements ISequentialLoader<WritableColumnVector> {
+public class SparkNullLoader implements ILoader<WritableColumnVector> {
 
   private final WritableColumnVector vector;
   private final int loadSize;
@@ -42,17 +42,22 @@ public int getLoadSize() {
 
   @Override
   public void setNull(final int index) throws IOException {
-    // TODO:
+    // FIXME:
   }
 
   @Override
   public void finish() throws IOException {
     // FIXME:
-    vector.putNulls(0, loadSize);
   }
 
   @Override
   public WritableColumnVector build() throws IOException {
+    vector.putNulls(0, loadSize);
     return vector;
   }
+
+  @Override
+  public boolean isLoadingSkipped() {
+    return true;
+  }
 }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkRunLengthEncodingArrayLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkRunLengthEncodingArrayLoader.java
index 9c5dda1..3ebb236 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkRunLengthEncodingArrayLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkRunLengthEncodingArrayLoader.java
@@ -49,7 +49,7 @@ public WritableColumnVector build() throws IOException {
 
   @Override
   public void setNull(final int index) throws IOException {
-    vector.putNull(index);
+    vector.putArray(index, 0, 0);
   }
 
   @Override
@@ -59,7 +59,7 @@ public void setRowGroupCount(final int count) throws IOException {}
   public void setNullAndRepetitions(
       final int startIndex, final int repetitions, final int rowGroupIndex) throws IOException {
     for (int i = 0; i < repetitions; i++) {
-      vector.putNull(rowId);
+      vector.putArray(rowId, 0, 0);
       rowId++;
     }
   }
@@ -83,6 +83,10 @@ public void setRowGourpIndexAndRepetitions(
   public void loadChild(final ColumnBinary columnBinary, final int childLength) throws IOException {
     vector.getChild(0).reset();
     vector.getChild(0).reserve(childLength);
+    if (vector.getChild(0).hasDictionary()) {
+      vector.getChild(0).reserveDictionaryIds(0);
+      vector.getChild(0).setDictionary(null);
+    }
     SparkLoaderFactoryUtil.createLoaderFactory(vector.getChild(0))
         .create(columnBinary, childLength);
   }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkStructLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkStructLoader.java
index fb2e52e..828a1c7 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkStructLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkStructLoader.java
@@ -70,7 +70,7 @@ public WritableColumnVector build() throws IOException {
     // NOTE: Fill unloaded columns with nulls.
     for (int i = 0; i < names.length; i++) {
       if (loaderFactoryMap.containsKey(names[i])) {
-        vector.getChild(i).putNulls(0, loadSize);
+        SparkEmptyLoader.load(vector.getChild(i), loadSize);
       }
     }
     return vector;
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionArrayLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionArrayLoader.java
index 346fe26..0950840 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionArrayLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionArrayLoader.java
@@ -29,6 +29,12 @@ public class SparkUnionArrayLoader implements IUnionLoader<WritableColumnVector>
   public SparkUnionArrayLoader(final WritableColumnVector vector, final int loadSize) {
     this.vector = vector;
     this.loadSize = loadSize;
+    this.vector.getChild(0).reset();
+    this.vector.getChild(0).reserve(0);
+    if (this.vector.getChild(0).hasDictionary()) {
+      this.vector.getChild(0).reserveDictionaryIds(0);
+      this.vector.getChild(0).setDictionary(null);
+    }
   }
 
   @Override
@@ -38,12 +44,12 @@ public int getLoadSize() {
 
   @Override
   public void setNull(final int index) throws IOException {
-    vector.putNull(index);
+    vector.putArray(index, 0, 0);
   }
 
   @Override
   public void finish() throws IOException {
-    //
+    // FIXME:
   }
 
   @Override
@@ -55,13 +61,14 @@ public WritableColumnVector build() throws IOException {
   public void setIndexAndColumnType(final int index, final ColumnType columnType) throws IOException {
     // FIXME:
     if (columnType != ColumnType.ARRAY) {
-      vector.putNull(index);
+      vector.putArray(index, 0, 0);
     }
   }
 
   @Override
   public void loadChild(final ColumnBinary columnBinary, final int childLoadSize) throws IOException {
     if (columnBinary.columnType == ColumnType.ARRAY) {
+      vector.getChild(0).reserve(childLoadSize);
       SparkLoaderFactoryUtil.createLoaderFactory(vector).create(columnBinary, childLoadSize);
     }
   }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionMapLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionMapLoader.java
index 107ccb0..68274e2 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionMapLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionMapLoader.java
@@ -29,6 +29,10 @@ public class SparkUnionMapLoader implements IUnionLoader<WritableColumnVector> {
   public SparkUnionMapLoader(WritableColumnVector vector, int loadSize) {
     this.vector = vector;
     this.loadSize = loadSize;
+    this.vector.getChild(0).reset();
+    this.vector.getChild(0).reserve(0);
+    this.vector.getChild(1).reset();
+    this.vector.getChild(1).reserve(0);
   }
 
   @Override
@@ -38,12 +42,12 @@ public int getLoadSize() {
 
   @Override
   public void setNull(int index) throws IOException {
-    vector.putNull(index);
+    // FIXME:
   }
 
   @Override
   public void finish() throws IOException {
-    //
+    // FIXME:
   }
 
   @Override
@@ -54,9 +58,6 @@ public WritableColumnVector build() throws IOException {
   @Override
   public void setIndexAndColumnType(int index, ColumnType columnType) throws IOException {
     // FIXME:
-    if (columnType != ColumnType.SPREAD) {
-      vector.putNull(index);
-    }
   }
 
   @Override
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionStructLoader.java b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionStructLoader.java
index c4e86de..dfc2341 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionStructLoader.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkUnionStructLoader.java
@@ -25,6 +25,7 @@
 public class SparkUnionStructLoader implements IUnionLoader<WritableColumnVector> {
   private final WritableColumnVector vector;
   private final int loadSize;
+  private boolean childLoaded;
 
   public SparkUnionStructLoader(WritableColumnVector vector, int loadSize) {
     this.vector = vector;
@@ -38,30 +39,31 @@ public int getLoadSize() {
 
   @Override
   public void setNull(int index) throws IOException {
-    vector.putNull(index);
+    // FIXME:
   }
 
   @Override
   public void finish() throws IOException {
-    //
+    // FIXME:
   }
 
   @Override
   public WritableColumnVector build() throws IOException {
+    if (!childLoaded) {
+      new SparkEmptyStructLoader(vector, loadSize).build();
+    }
     return vector;
   }
 
   @Override
   public void setIndexAndColumnType(int index, ColumnType columnType) throws IOException {
     // FIXME:
-    if (columnType != ColumnType.SPREAD) {
-      vector.putNull(index);
-    }
   }
 
   @Override
   public void loadChild(ColumnBinary columnBinary, int childLoadSize) throws IOException {
     if (columnBinary.columnType == ColumnType.SPREAD) {
+      childLoaded = true;
       SparkLoaderFactoryUtil.createLoaderFactory(vector).create(columnBinary, childLoadSize);
     }
   }
diff --git a/src/main/java/jp/co/yahoo/yosegi/spark/reader/SparkColumnarBatchConverter.java b/src/main/java/jp/co/yahoo/yosegi/spark/reader/SparkColumnarBatchConverter.java
index fb38e8f..19b9a58 100644
--- a/src/main/java/jp/co/yahoo/yosegi/spark/reader/SparkColumnarBatchConverter.java
+++ b/src/main/java/jp/co/yahoo/yosegi/spark/reader/SparkColumnarBatchConverter.java
@@ -17,9 +17,11 @@
 import jp.co.yahoo.yosegi.binary.ColumnBinary;
 import jp.co.yahoo.yosegi.inmemory.IRawConverter;
 import jp.co.yahoo.yosegi.spark.inmemory.SparkLoaderFactoryUtil;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyLoader;
 import jp.co.yahoo.yosegi.spark.utils.PartitionColumnUtil;
 import org.apache.spark.sql.catalyst.InternalRow;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.ArrayType;
 import org.apache.spark.sql.types.StructType;
 import org.apache.spark.sql.vectorized.ColumnVector;
 import org.apache.spark.sql.vectorized.ColumnarBatch;
@@ -72,10 +74,10 @@ public ColumnarBatch convert(final List<ColumnBinary> raw, final int loadSize) t
       isSet[index] = true;
       SparkLoaderFactoryUtil.createLoaderFactory(((WritableColumnVector) childColumns[index])).create(columnBinary, loadSize);
     }
-    // NOTE: null columns
+    // NOTE: Empty columns
     for (int i = 0; i < schema.length(); i++) {
       if (!isSet[i]) {
-        ((WritableColumnVector) childColumns[i]).putNulls(0, loadSize);
+        SparkEmptyLoader.load((WritableColumnVector) childColumns[i], loadSize);
       }
     }
     // NOTE: partitionColumns
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/blackbox/LoadTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/blackbox/LoadTest.java
index 0b8b6c6..65ddaf7 100644
--- a/src/test/java/jp/co/yahoo/yosegi/spark/blackbox/LoadTest.java
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/blackbox/LoadTest.java
@@ -322,33 +322,24 @@ void T_load_Array_Array_Integer_1() throws IOException {
       final int jIndex = ldfj.get(i).fieldIndex("aa");
       final int yIndex = ldfy.get(i).fieldIndex("aa");
       if (ldfj.get(i).isNullAt(jIndex)) {
-        if (ldfy.get(i).isNullAt(yIndex)) {
-          // NOTE: json:null, yosegi:null
-          assertTrue(ldfy.get(i).isNullAt(yIndex));
-        } else {
-          // FIXME: json:null, yosegi:[]
-          assertTrue(false);
-        }
+        // NOTE: json:null, yosegi:[]
+        assertEquals(0, ldfy.get(i).getList(yIndex).size());
       } else {
         if (ldfy.get(i).isNullAt(yIndex)) {
           // NOTE: json:[], yosegi:null
-          assertEquals(0, ldfj.get(i).getList(jIndex).size());
+          assertTrue(false);
         } else {
           final List<WrappedArray<Long>> ldfj2 = ldfj.get(i).getList(jIndex);
           final List<WrappedArray<Long>> ldfy2 = ldfy.get(i).getList(yIndex);
           for (int j = 0; j < ldfj2.size(); j++) {
             final WrappedArray<Long> waj = ldfj2.get(j);
             final WrappedArray<Long> way = ldfy2.get(j);
-            if (way == null) {
-              if (waj == null) {
-                // NOTE: json:[null], yosegi:[null]
-                assertNull(waj);
-              } else {
-                // NOTE: json:[[]], yosegi:[null]
-                assertEquals(0, waj.size());
-              }
+            if (waj == null) {
+              // NOTE: json:[null], yosegi:[[]]
+              assertEquals(0, way.size());
             } else {
               // NOTE: json:[[]], yosegi:[[]]
+              assertEquals(waj.size(), way.size());
               for (int k = 0; k < waj.size(); k++) {
                 assertEquals(waj.apply(k), way.apply(k));
               }
@@ -397,59 +388,28 @@ void T_load_Array_Struct_Primitive_1() throws IOException {
       final int jIndex = ldfj.get(i).fieldIndex("as");
       final int yIndex = ldfy.get(i).fieldIndex("as");
       if (ldfj.get(i).isNullAt(jIndex)) {
-        if (ldfy.get(i).isNullAt(yIndex)) {
-          // NOTE: json:null, yosegi:null
-          assertTrue(ldfy.get(i).isNullAt(yIndex));
-        } else {
-          // FIXME: json:null, yosegi:[]
-          assertTrue(false);
-        }
+        // NOTE: json:null, yosegi:[]
+        assertEquals(0, ldfy.get(i).getList(yIndex).size());
       } else {
         if (ldfy.get(i).isNullAt(yIndex)) {
-          final List<Row> lrj = ldfj.get(i).getList(jIndex);
-          for (int j = 0; j < lrj.size(); j++) {
-            final Row rj = lrj.get(j);
-            if (rj == null) {
-              // NOTE: json:[null], yosegi:null
-              assertNull(rj);
-            } else {
-              // NOTE: json[as.field:null], yosegi:null
-              for (final StructField field : fields) {
-                final String name = field.name();
-                assertNull(rj.getAs(name));
-              }
-            }
-          }
+          assertTrue(false);
         } else {
           final List<Row> lrj = ldfj.get(i).getList(jIndex);
           final List<Row> lry = ldfy.get(i).getList(yIndex);
           for (int j = 0; j < lrj.size(); j++) {
             final Row rj = lrj.get(j);
             final Row ry = lry.get(j);
-            if (ry == null) {
-              if (rj == null) {
-                // NOTE: json:[null], yosegi:[null]
-                assertNull(rj);
-              } else {
-                // NOTE: json:[{}], yosegi:[null]
-                for (final StructField field : fields) {
-                  final String name = field.name();
-                  assertNull(rj.getAs(name));
-                }
+            if (rj == null) {
+              // NOTE: json:[null], yosegi:[{}]
+              for (final StructField field : fields) {
+                final String name = field.name();
+                assertNull(ry.getAs(name));
               }
             } else {
-              if (rj == null) {
-                // NOTE: json:[null], yosegi:[{}]
-                for (final StructField field : fields) {
-                  final String name = field.name();
-                  assertNull(ry.getAs(name));
-                }
-              } else {
-                // NOTE: json:[{}], yosegi:[{}]
-                for (final StructField field : fields) {
-                  final String name = field.name();
-                  assertEquals((Object) rj.getAs(name), (Object) ry.getAs(name));
-                }
+              // NOTE: json:[{}], yosegi:[{}]
+              for (final StructField field : fields) {
+                final String name = field.name();
+                assertEquals((Object) rj.getAs(name), (Object) ry.getAs(name));
               }
             }
           }
@@ -484,31 +444,12 @@ void T_load_Array_Map_Integer_1() throws IOException {
       final int jIndex = ldfj.get(i).fieldIndex("am");
       final int yIndex = ldfy.get(i).fieldIndex("am");
       if (ldfj.get(i).isNullAt(jIndex)) {
-        if (ldfy.get(i).isNullAt(yIndex)) {
-          // NOTE: json:null, yosegi:null
-          assertTrue(ldfy.get(i).isNullAt(yIndex));
-        } else {
-          // FIXME: json:null, yosegi:[]
-          assertTrue(false);
-        }
+        // NOTE: json:null, yosegi:[]
+        assertEquals(0, ldfy.get(i).getList(yIndex).size());
       } else {
         if (ldfy.get(i).isNullAt(yIndex)) {
           // NOTE: json:[], yosegi:null
-          final List<Map<String, Integer>> lmj = ldfj.get(i).getList(jIndex);
-          for (int j = 0; j < lmj.size(); j++) {
-            final Map<String, Integer> mj = lmj.get(j);
-            if (mj == null) {
-              // NOTE: json:[null], yosegi:null
-              assertNull(mj);
-            } else {
-              // NOTE: json:[am.key:null], yosegi:null
-              final Iterator<String> iter = mj.keysIterator();
-              while (iter.hasNext()) {
-                final String key = iter.next();
-                assertNull(mj.get(key).get());
-              }
-            }
-          }
+          assertTrue(false);
         } else {
           // NOTE: json:[], yosegi:[]
           final List<Map<String, Integer>> lmj = ldfj.get(i).getList(jIndex);
@@ -516,38 +457,20 @@ void T_load_Array_Map_Integer_1() throws IOException {
           for (int j = 0; j < lmj.size(); j++) {
             final Map<String, Integer> mj = lmj.get(j);
             final Map<String, Integer> my = lmy.get(j);
-            if (my == null) {
-              if (mj == null) {
-                // NOTE: json:[null], yosegi:[null]
-                assertNull(mj);
-              } else {
-                // NOTE: json:[{}], yosegi:[null]
-                final Iterator<String> iter = mj.keysIterator();
-                while (iter.hasNext()) {
-                  final String key = iter.next();
-                  assertNull(mj.get(key).get());
-                }
-              }
+            if (mj == null) {
+              // NOTE: json:[null], yosegi:[{}]
+              assertEquals(0, my.size());
             } else {
-              if (mj == null) {
-                // NOTE: json:[null], yosegi:[{}]
-                final Iterator<String> iter = my.keysIterator();
-                while (iter.hasNext()) {
-                  final String key = iter.next();
-                  assertNull(my.get(key).get());
-                }
-              } else {
-                // NOTE: json[{}], yosegi:[{}]
-                final Iterator<String> iter = mj.keysIterator();
-                while (iter.hasNext()) {
-                  final String key = iter.next();
-                  if (mj.get(key).get() == null) {
-                    // NOTE: json:[{key:null}], yosegi:[{key:not exist}]
-                    assertTrue(my.get(key).isEmpty());
-                  } else {
-                    // NOTE: json:[{key}], yosegi:[{key}]
-                    assertEquals(mj.get(key), my.get(key));
-                  }
+              // NOTE: json[{}], yosegi:[{}]
+              final Iterator<String> iter = mj.keysIterator();
+              while (iter.hasNext()) {
+                final String key = iter.next();
+                if (mj.get(key).get() == null) {
+                  // NOTE: json:[{key:null}], yosegi:[{key:not exist}]
+                  assertTrue(my.get(key).isEmpty());
+                } else {
+                  // NOTE: json:[{key}], yosegi:[{key}]
+                  assertEquals(mj.get(key), my.get(key));
                 }
               }
             }
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactoryTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactoryTest.java
index 7cab7a4..0dbab82 100644
--- a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactoryTest.java
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkArrayLoaderFactoryTest.java
@@ -19,7 +19,7 @@
 import jp.co.yahoo.yosegi.binary.maker.MaxLengthBasedArrayColumnBinaryMaker;
 import jp.co.yahoo.yosegi.inmemory.ILoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkArrayLoader;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyArrayLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkRunLengthEncodingArrayLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkUnionArrayLoader;
 import jp.co.yahoo.yosegi.spark.test.Utils;
@@ -99,6 +99,6 @@ void T_createLoader_Null() throws IOException {
     final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
     final ILoader loader = new SparkArrayLoaderFactory(vector).createLoader(columnBinary, loadSize);
 
-    assertTrue(loader instanceof SparkNullLoader);
+    assertTrue(loader instanceof SparkEmptyArrayLoader);
   }
 }
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactoryTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactoryTest.java
index 21b2f8e..1a823fe 100644
--- a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactoryTest.java
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkMapLoaderFactoryTest.java
@@ -17,8 +17,8 @@
 import jp.co.yahoo.yosegi.binary.ColumnBinary;
 import jp.co.yahoo.yosegi.binary.maker.DumpSpreadColumnBinaryMaker;
 import jp.co.yahoo.yosegi.inmemory.ILoader;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyMapLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkMapLoader;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
 import jp.co.yahoo.yosegi.spark.test.Utils;
 import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
@@ -64,6 +64,6 @@ void T_createLoader_Null() throws IOException {
     final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
     final ILoader loader = new SparkMapLoaderFactory(vector).createLoader(columnBinary, loadSize);
 
-    assertTrue(loader instanceof SparkNullLoader);
+    assertTrue(loader instanceof SparkEmptyMapLoader);
   }
 }
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactoryTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactoryTest.java
index 1a458fd..1e66d38 100644
--- a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactoryTest.java
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/factory/SparkStructLoaderFactoryTest.java
@@ -17,7 +17,7 @@
 import jp.co.yahoo.yosegi.binary.ColumnBinary;
 import jp.co.yahoo.yosegi.binary.maker.DumpSpreadColumnBinaryMaker;
 import jp.co.yahoo.yosegi.inmemory.ILoader;
-import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkNullLoader;
+import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkEmptyStructLoader;
 import jp.co.yahoo.yosegi.spark.inmemory.loader.SparkStructLoader;
 import jp.co.yahoo.yosegi.spark.test.Utils;
 import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
@@ -89,6 +89,6 @@ void T_createLoader_Null() throws IOException {
     final ILoader loader =
         new SparkStructLoaderFactory(vector).createLoader(columnBinary, loadSize);
 
-    assertTrue(loader instanceof SparkNullLoader);
+    assertTrue(loader instanceof SparkEmptyStructLoader);
   }
 }
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoaderTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoaderTest.java
new file mode 100644
index 0000000..db61a1a
--- /dev/null
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyArrayLoaderTest.java
@@ -0,0 +1,104 @@
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.params.ParameterizedTest;
+import org.junit.jupiter.params.provider.Arguments;
+import org.junit.jupiter.params.provider.MethodSource;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.List;
+import java.util.stream.Stream;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.params.provider.Arguments.arguments;
+
+class SparkEmptyArrayLoaderTest {
+    public static Stream<Arguments> data() {
+        final int loadSize = 5;
+        return Stream.of(
+            arguments(DataTypes.BooleanType, loadSize),
+            arguments(DataTypes.ByteType, loadSize),
+            arguments(DataTypes.StringType, loadSize),
+            arguments(DataTypes.createDecimalType(), loadSize),
+            arguments(DataTypes.DoubleType, loadSize),
+            arguments(DataTypes.FloatType, loadSize),
+            arguments(DataTypes.IntegerType, loadSize),
+            arguments(DataTypes.LongType, loadSize),
+            arguments(DataTypes.ShortType, loadSize)
+        );
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Primitive(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType dataType = DataTypes.createArrayType(elmDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyArrayLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Array(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType arrayDataType = DataTypes.createArrayType(elmDataType);
+        final DataType dataType = DataTypes.createArrayType(arrayDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyArrayLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Map(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType mapDataType = DataTypes.createMapType(DataTypes.StringType, elmDataType, true);
+        final DataType dataType = DataTypes.createArrayType(mapDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyArrayLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @Test
+    void T_load_Array_Struct() throws IOException {
+        final List<StructField> fields =
+            Arrays.asList(
+                DataTypes.createStructField("bo", DataTypes.BooleanType, true),
+                DataTypes.createStructField("by", DataTypes.ByteType, true),
+                DataTypes.createStructField("bi", DataTypes.BinaryType, true),
+                DataTypes.createStructField("do", DataTypes.DoubleType, true),
+                DataTypes.createStructField("fl", DataTypes.FloatType, true),
+                DataTypes.createStructField("in", DataTypes.IntegerType, true),
+                DataTypes.createStructField("lo", DataTypes.LongType, true),
+                DataTypes.createStructField("sh", DataTypes.ShortType, true),
+                DataTypes.createStructField("st", DataTypes.StringType, true));
+        final DataType structDataType = DataTypes.createStructType(fields);
+        final DataType dataType = DataTypes.createArrayType(structDataType, true);
+        final int loadSize = 5;
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyArrayLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoaderTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoaderTest.java
new file mode 100644
index 0000000..b7a23f4
--- /dev/null
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyLoaderTest.java
@@ -0,0 +1,186 @@
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.vectorized.ColumnarMap;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.params.ParameterizedTest;
+import org.junit.jupiter.params.provider.Arguments;
+import org.junit.jupiter.params.provider.MethodSource;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.List;
+import java.util.stream.Stream;
+
+import static org.junit.jupiter.api.Assertions.*;
+import static org.junit.jupiter.params.provider.Arguments.arguments;
+
+class SparkEmptyLoaderTest {
+    public static Stream<Arguments> data() {
+        final int loadSize = 5;
+        return Stream.of(
+            arguments(DataTypes.BooleanType, loadSize),
+            arguments(DataTypes.ByteType, loadSize),
+            arguments(DataTypes.StringType, loadSize),
+            arguments(DataTypes.createDecimalType(), loadSize),
+            arguments(DataTypes.DoubleType, loadSize),
+            arguments(DataTypes.FloatType, loadSize),
+            arguments(DataTypes.IntegerType, loadSize),
+            arguments(DataTypes.LongType, loadSize),
+            arguments(DataTypes.ShortType, loadSize)
+        );
+    }
+
+    public void assertNull(final WritableColumnVector vector, final int loadSize) {
+        for (int i = 0; i < loadSize; i++) {
+            assertTrue(vector.isNullAt(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Primitive(final DataType dataType, final int loadSize) throws IOException {
+        final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        assertNull(vector, loadSize);
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Primitive(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType dataType = DataTypes.createArrayType(elmDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Array(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType arrayDataType = DataTypes.createArrayType(elmDataType);
+        final DataType dataType = DataTypes.createArrayType(arrayDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Array_Map(final DataType elmDataType, final int loadSize) throws IOException {
+        final DataType mapDataType = DataTypes.createMapType(DataTypes.StringType, elmDataType, true);
+        final DataType dataType = DataTypes.createArrayType(mapDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @Test
+    void T_load_Array_Struct() throws IOException {
+        final List<StructField> fields =
+            Arrays.asList(
+                DataTypes.createStructField("bo", DataTypes.BooleanType, true),
+                DataTypes.createStructField("by", DataTypes.ByteType, true),
+                DataTypes.createStructField("bi", DataTypes.BinaryType, true),
+                DataTypes.createStructField("do", DataTypes.DoubleType, true),
+                DataTypes.createStructField("fl", DataTypes.FloatType, true),
+                DataTypes.createStructField("in", DataTypes.IntegerType, true),
+                DataTypes.createStructField("lo", DataTypes.LongType, true),
+                DataTypes.createStructField("sh", DataTypes.ShortType, true),
+                DataTypes.createStructField("st", DataTypes.StringType, true));
+        final DataType structDataType = DataTypes.createStructType(fields);
+        final DataType dataType = DataTypes.createArrayType(structDataType, true);
+        final int loadSize = 5;
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            assertEquals(0, vector.getArrayLength(i));
+        }
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load_Map(final DataType elmDataType, final int loadSize) throws Exception {
+        final DataType dataType = DataTypes.createMapType(DataTypes.StringType, elmDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            final ColumnarMap cm = vector.getMap(i);
+            assertEquals(0, cm.numElements());
+        }
+    }
+
+    @Test
+    void T_load_Struct() throws IOException {
+        final List<StructField> e4Fields =
+            Arrays.asList(
+                DataTypes.createStructField("e1", DataTypes.LongType, true),
+                DataTypes.createStructField("e2", DataTypes.createArrayType(DataTypes.LongType, true), true),
+                DataTypes.createStructField("e3",
+                    DataTypes.createMapType(DataTypes.StringType, DataTypes.LongType, true), true)
+            );
+        final List<StructField> fields =
+            Arrays.asList(
+                DataTypes.createStructField("e1", DataTypes.StringType, true),
+                DataTypes.createStructField("e2", DataTypes.createArrayType(DataTypes.StringType, true), true),
+                DataTypes.createStructField("e3",
+                    DataTypes.createMapType(DataTypes.StringType, DataTypes.StringType, true), true),
+                DataTypes.createStructField("e4", DataTypes.createStructType(e4Fields), true)
+            );
+        final DataType dataType = DataTypes.createStructType(fields);
+        final int loadSize = 5;
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        SparkEmptyLoader.load(vector, loadSize);
+        for (int i = 0; i < loadSize; i++) {
+            for (int j = 0; j < fields.size(); j++) {
+                final StructField field = fields.get(j);
+                final String name = field.name();
+                final DataType type = field.dataType();
+                if (name.equals("e1")) {
+                    assertTrue(vector.getChild(j).isNullAt(i));
+                } else if (name.equals("e2")) {
+                    assertFalse(vector.getChild(j).isNullAt(i));
+                    assertEquals(0, vector.getChild(j).getArrayLength(i));
+                } else if (name.equals("e3")) {
+                    assertFalse(vector.getChild(j).isNullAt(i));
+                    final ColumnarMap cm = vector.getChild(j).getMap(i);
+                    assertEquals(0, cm.numElements());
+                } else if (name.equals("e4")) {
+                    for (int k = 0; k < e4Fields.size(); k++) {
+                        final StructField e4Field = e4Fields.get(k);
+                        final String e4Name = e4Field.name();
+                        final DataType e4Type = e4Field.dataType();
+                        if (e4Name.equals("e1")) {
+                            assertTrue(vector.getChild(j).getChild(k).isNullAt(i));
+                        } else if (e4Name.equals("e2")) {
+                            assertFalse(vector.getChild(j).isNullAt(i));
+                            assertEquals(0, vector.getChild(j).getChild(k).getArrayLength(i));
+                        } else if (e4Name.equals("e3")) {
+                            assertFalse(vector.getChild(j).isNullAt(i));
+                            final ColumnarMap e4Cm = vector.getChild(j).getChild(k).getMap(i);
+                            assertEquals(0, e4Cm.numElements());
+                        } else {
+                            assertTrue(false);
+                        }
+                    }
+                } else {
+                    assertTrue(false);
+                }
+            }
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoaderTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoaderTest.java
new file mode 100644
index 0000000..208a377
--- /dev/null
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyMapLoaderTest.java
@@ -0,0 +1,48 @@
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.vectorized.ColumnarMap;
+import org.junit.jupiter.params.ParameterizedTest;
+import org.junit.jupiter.params.provider.Arguments;
+import org.junit.jupiter.params.provider.MethodSource;
+
+import java.util.stream.Stream;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.params.provider.Arguments.arguments;
+
+class SparkEmptyMapLoaderTest {
+    public static Stream<Arguments> data() {
+        final int loadSize = 5;
+        return Stream.of(
+            arguments(DataTypes.BooleanType, loadSize),
+            arguments(DataTypes.ByteType, loadSize),
+            arguments(DataTypes.StringType, loadSize),
+            arguments(DataTypes.createDecimalType(), loadSize),
+            arguments(DataTypes.DoubleType, loadSize),
+            arguments(DataTypes.FloatType, loadSize),
+            arguments(DataTypes.IntegerType, loadSize),
+            arguments(DataTypes.LongType, loadSize),
+            arguments(DataTypes.ShortType, loadSize)
+        );
+    }
+
+    @ParameterizedTest
+    @MethodSource("data")
+    void T_load(final DataType elmDataType, final int loadSize) throws Exception {
+        final DataType dataType = DataTypes.createMapType(DataTypes.StringType, elmDataType, true);
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyMapLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            assertFalse(vector.isNullAt(i));
+            final ColumnarMap cm = vector.getMap(i);
+            assertEquals(0, cm.numElements());
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoaderTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoaderTest.java
new file mode 100644
index 0000000..869e167
--- /dev/null
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkEmptyStructLoaderTest.java
@@ -0,0 +1,81 @@
+package jp.co.yahoo.yosegi.spark.inmemory.loader;
+
+import jp.co.yahoo.yosegi.inmemory.ILoader;
+import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
+import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.vectorized.ColumnarMap;
+import org.junit.jupiter.api.Test;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.List;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+class SparkEmptyStructLoaderTest {
+    @Test
+    void T_load() throws IOException {
+        final List<StructField> e4Fields =
+            Arrays.asList(
+                DataTypes.createStructField("e1", DataTypes.LongType, true),
+                DataTypes.createStructField("e2", DataTypes.createArrayType(DataTypes.LongType, true), true),
+                DataTypes.createStructField("e3",
+                    DataTypes.createMapType(DataTypes.StringType, DataTypes.LongType, true), true)
+            );
+        final List<StructField> fields =
+            Arrays.asList(
+                DataTypes.createStructField("e1", DataTypes.StringType, true),
+                DataTypes.createStructField("e2", DataTypes.createArrayType(DataTypes.StringType, true), true),
+                DataTypes.createStructField("e3",
+                    DataTypes.createMapType(DataTypes.StringType, DataTypes.StringType, true), true),
+                DataTypes.createStructField("e4", DataTypes.createStructType(e4Fields), true)
+            );
+        final DataType dataType = DataTypes.createStructType(fields);
+        final int loadSize = 5;
+        final WritableColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
+        final ILoader<WritableColumnVector> loader = new SparkEmptyStructLoader(vector, loadSize);
+        loader.build();
+        for (int i = 0; i < loadSize; i++) {
+            for (int j = 0; j < fields.size(); j++) {
+                final StructField field = fields.get(j);
+                final String name = field.name();
+                final DataType type = field.dataType();
+                if (name.equals("e1")) {
+                    assertTrue(vector.getChild(j).isNullAt(i));
+                } else if (name.equals("e2")) {
+                    assertFalse(vector.getChild(j).isNullAt(i));
+                    assertEquals(0, vector.getChild(j).getArrayLength(i));
+                } else if (name.equals("e3")) {
+                    assertFalse(vector.getChild(j).isNullAt(i));
+                    final ColumnarMap cm = vector.getChild(j).getMap(i);
+                    assertEquals(0, cm.numElements());
+                } else if (name.equals("e4")) {
+                    for (int k = 0; k < e4Fields.size(); k++) {
+                        final StructField e4Field = e4Fields.get(k);
+                        final String e4Name = e4Field.name();
+                        final DataType e4Type = e4Field.dataType();
+                        if (e4Name.equals("e1")) {
+                            assertTrue(vector.getChild(j).getChild(k).isNullAt(i));
+                        } else if (e4Name.equals("e2")) {
+                            assertFalse(vector.getChild(j).isNullAt(i));
+                            assertEquals(0, vector.getChild(j).getChild(k).getArrayLength(i));
+                        } else if (e4Name.equals("e3")) {
+                            assertFalse(vector.getChild(j).isNullAt(i));
+                            final ColumnarMap e4Cm = vector.getChild(j).getChild(k).getMap(i);
+                            assertEquals(0, e4Cm.numElements());
+                        } else {
+                            assertTrue(false);
+                        }
+                    }
+                } else {
+                    assertTrue(false);
+                }
+            }
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoaderTest.java b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoaderTest.java
index 6893be5..86ae384 100644
--- a/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoaderTest.java
+++ b/src/test/java/jp/co/yahoo/yosegi/spark/inmemory/loader/SparkNullLoaderTest.java
@@ -14,7 +14,7 @@
  */
 package jp.co.yahoo.yosegi.spark.inmemory.loader;
 
-import jp.co.yahoo.yosegi.inmemory.ISequentialLoader;
+import jp.co.yahoo.yosegi.inmemory.ILoader;
 import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector;
 import org.apache.spark.sql.execution.vectorized.WritableColumnVector;
 import org.apache.spark.sql.types.DataType;
@@ -38,8 +38,8 @@ void T_load_Boolean_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.BooleanType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -51,8 +51,8 @@ void T_load_Byte_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.ByteType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -64,8 +64,8 @@ void T_load_Bytes_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.BinaryType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -77,8 +77,8 @@ void T_load_Double_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.DoubleType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -90,8 +90,8 @@ void T_load_Float_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.FloatType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -103,8 +103,8 @@ void T_load_Integer_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.IntegerType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -116,8 +116,8 @@ void T_load_Long_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.LongType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -129,8 +129,8 @@ void T_load_Short_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.ShortType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);
@@ -142,8 +142,8 @@ void T_load_String_1() throws IOException {
     final int loadSize = 5;
     final DataType dataType = DataTypes.StringType;
     final OnHeapColumnVector vector = new OnHeapColumnVector(loadSize, dataType);
-    final ISequentialLoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
-    loader.finish();
+    final ILoader<WritableColumnVector> loader = new SparkNullLoader(vector, loadSize);
+    loader.build();
 
     // NOTE: assert
     assertNull(vector, loadSize);