From 19c41464c768bbf01b183a7466dff37c20ab1d36 Mon Sep 17 00:00:00 2001
From: WeichenXu <weichen.xu@databricks.com>
Date: Sat, 2 Sep 2017 04:51:05 +0800
Subject: [PATCH] fix df.drop in VectorAssembler (#117)

fix df.drop in VectorAssembler to return correct DataFrame
---
 .../spark/sql/perf/mllib/feature/VectorAssembler.scala           | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/main/scala/com/databricks/spark/sql/perf/mllib/feature/VectorAssembler.scala b/src/main/scala/com/databricks/spark/sql/perf/mllib/feature/VectorAssembler.scala
index 914abec..7e862b5 100644
--- a/src/main/scala/com/databricks/spark/sql/perf/mllib/feature/VectorAssembler.scala
+++ b/src/main/scala/com/databricks/spark/sql/perf/mllib/feature/VectorAssembler.scala
@@ -34,7 +34,6 @@ object VectorAssembler extends BenchmarkAlgorithm with TestFromTraining {
       df = df.withColumn(colName, sliceVec(col("features"), lit(fromIndex), lit(untilIndex)))
     }
     df.drop(col("features"))
-    df
   }
 
   override def getPipelineStage(ctx: MLBenchContext): PipelineStage = {