Dataset API: Allow empty projection (#24)

Author: zhztheplayer

cpp/src/jni/dataset/jni_wrapper.cc

@@ -590,9 +590,7 @@ JNIEXPORT jlong JNICALL Java_org_apache_arrow_dataset_jni_JniWrapper_createScann
   JniAssertOkOrThrow(scanner_builder->Pool(pool));
 
   std::vector<std::string> column_vector = ToStringVector(env, columns);
-  if (!column_vector.empty()) {
-    JniAssertOkOrThrow(scanner_builder->Project(column_vector));
-  }
+  JniAssertOkOrThrow(scanner_builder->Project(column_vector));
   JniAssertOkOrThrow(scanner_builder->BatchSize(batch_size));
   // initialize filters
   jsize exprs_len = env->GetArrayLength(filter);

java/dataset/src/test/java/org/apache/arrow/dataset/file/TestFileSystemDataset.java

@@ -137,6 +137,27 @@ public void testParquetProjector() throws Exception {
     AutoCloseables.close(factory);
   }
 
+  @Test
+  public void testEmptyProjector() throws Exception {
+    ParquetWriteSupport writeSupport = ParquetWriteSupport.writeTempFile(AVRO_SCHEMA_USER, TMP.newFolder(), 1, "a",
+            2, "b");
+
+    FileSystemDatasetFactory factory = new FileSystemDatasetFactory(rootAllocator(), NativeMemoryPool.getDefault(),
+            FileFormat.PARQUET, writeSupport.getOutputURI());
+    ScanOptions options = new ScanOptions(new String[]{}, Filter.EMPTY, 100);
+    Schema schema = inferResultSchemaFromFactory(factory, options);
+    List<ArrowRecordBatch> datum = collectResultFromFactory(factory, options);
+    org.apache.avro.Schema expectedSchema = truncateAvroSchema(writeSupport.getAvroSchema(), 0, 1);
+
+    assertSingleTaskProduced(factory, options);
+    assertEquals(0, schema.getFields().size());
+    assertEquals(1, datum.size());
+
+    AutoCloseables.close(datum);
+    AutoCloseables.close(factory);
+  }
+
+
   @Test
   public void testParquetBatchSize() throws Exception {
     ParquetWriteSupport writeSupport = ParquetWriteSupport.writeTempFile(AVRO_SCHEMA_USER, TMP.newFolder(),