From b08673475c901caf78b938bad10a14a920c7ec56 Mon Sep 17 00:00:00 2001
From: Yang Zhang <yangchuan.zy@alibaba-inc.com>
Date: Thu, 15 Aug 2024 14:36:27 +0800
Subject: [PATCH] [VL] Verify empty2null is offloaded when v1writer fallback
 (#6859)

---
 .../spark/sql/sources/GlutenInsertSuite.scala | 32 ++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)

diff --git a/gluten-ut/spark34/src/test/scala/org/apache/spark/sql/sources/GlutenInsertSuite.scala b/gluten-ut/spark34/src/test/scala/org/apache/spark/sql/sources/GlutenInsertSuite.scala
index 5c60115c5e1d..ca0ada39ceec 100644
--- a/gluten-ut/spark34/src/test/scala/org/apache/spark/sql/sources/GlutenInsertSuite.scala
+++ b/gluten-ut/spark34/src/test/scala/org/apache/spark/sql/sources/GlutenInsertSuite.scala
@@ -16,7 +16,7 @@
  */
 package org.apache.spark.sql.sources
 
-import org.apache.gluten.execution.SortExecTransformer
+import org.apache.gluten.execution.{ProjectExecTransformer, SortExecTransformer}
 import org.apache.gluten.extension.GlutenPlan
 
 import org.apache.spark.SparkConf
@@ -147,6 +147,36 @@ class GlutenInsertSuite
     assert(parts == expectedPartitionNames)
   }
 
+  testGluten("offload empty2null when v1writes fallback") {
+    withSQLConf((SQLConf.MAX_RECORDS_PER_FILE.key, "1000")) {
+      withTable("pt") {
+        spark.sql("CREATE TABLE pt (c1 int) USING PARQUET PARTITIONED BY(p string)")
+
+        val df = spark.sql(s"""
+                              |INSERT OVERWRITE TABLE pt PARTITION(p)
+                              |SELECT c1, c2 as p FROM source
+                              |""".stripMargin)
+
+        val writeFiles = stripAQEPlan(
+          df.queryExecution.executedPlan
+            .asInstanceOf[CommandResultExec]
+            .commandPhysicalPlan).children.head
+        assert(!writeFiles.isInstanceOf[ColumnarWriteFilesExec])
+        assert(writeFiles.exists(_.isInstanceOf[ProjectExecTransformer]))
+        val projectExecTransformer = writeFiles
+          .find(_.isInstanceOf[ProjectExecTransformer])
+          .get
+          .asInstanceOf[ProjectExecTransformer]
+        projectExecTransformer.projectList.find(_.toString().contains("empty2null"))
+
+        // The partition column should never be empty
+        checkAnswer(
+          spark.sql("SELECT * FROM pt"),
+          spark.sql("SELECT c1, if(c2 = '', null, c2) FROM source"))
+      }
+    }
+  }
+
   testGluten("remove v1writes sort and project") {
     // Only string type has empty2null expression
     withTable("pt") {