[GitHub] [spark] cloud-fan commented on a change in pull request #31325: [SPARK-34227][SQL] WindowFunctionFrame should clear its states during preparation

[GitBox <gi...@apache.org>]

cloud-fan commented on a change in pull request #31325:
URL: https://github.com/apache/spark/pull/31325#discussion_r563808912



##########
File path: sql/core/src/test/scala/org/apache/spark/sql/DataFrameWindowFunctionsSuite.scala
##########
@@ -1044,4 +1044,30 @@ class DataFrameWindowFunctionsSuite extends QueryTest
         Row(Seq(-0.0f, 0.0f), Row(-0.0d, Double.NaN), Seq(Row(-0.0d, Double.NaN)), 2),
         Row(Seq(0.0f, -0.0f), Row(0.0d, Double.NaN), Seq(Row(0.0d, 0.0/0.0)), 2)))
   }
+
+  test("SPARK-34227: WindowFunctionFrame should clear its states during preparation") {
+    // This creates a single partition dataframe with 3 records:
+    //   "a", 0, null
+    //   "a", 1, "x"
+    //   "b", 0, null
+    val df = spark.range(0, 3, 1, 1).select(
+      when($"id" < 2, lit("a")).otherwise(lit("b")).as("key"),
+      ($"id" % 2).cast("int").as("order"),
+      when($"id" % 2 === 0, lit(null)).otherwise(lit("x")).as("value"))
+
+    val window1 = Window.partitionBy($"key").orderBy($"order")
+      .rowsBetween(Window.unboundedPreceding, Window.unboundedFollowing)
+    val window2 = Window.partitionBy($"key").orderBy($"order")
+      .rowsBetween(Window.unboundedPreceding, Window.currentRow)
+    checkAnswer(
+      df.select(
+        $"key",
+        $"order",
+        nth_value($"value", 1, ignoreNulls = true).over(window1),
+        nth_value($"value", 1, ignoreNulls = true).over(window2)),
+      Seq(
+        Row("a", 0, "x", null),
+        Row("a", 1, "x", "x"),
+        Row("b", 0, null, null)))

Review comment:
       without this fix, when processing the next partition `"b"`, states are not reset and function still returns `"x"`




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org