cloud-fan commented on a change in pull request #27728:
[SPARK-25556][SPARK-17636][SPARK-31026][SPARK-31060][SQL][test-hive1.2] Nested
Column Predicate Pushdown for Parquet
URL: https://github.com/apache/spark/pull/27728#discussion_r397657185
##########
File path:
sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetFilterSuite.scala
##########
@@ -121,43 +121,81 @@ abstract class ParquetFilterSuite extends QueryTest with
ParquetTest with Shared
checkBinaryFilterPredicate(predicate, filterClass, Seq(Row(expected)))(df)
}
+ /**
+ * Takes single level `inputDF` dataframe to generate multi-level nested
+ * dataframes as new test data.
+ */
+ private def withNestedDataFrame(inputDF: DataFrame)
+ (testCases: (DataFrame, String, Any => Any) => Unit): Unit = {
+ val df = inputDF.toDF("temp")
+ Seq(
+ (
+ df.withColumnRenamed("temp", "a"),
+ "a", // zero nesting
+ (x: Any) => x),
+ (
+ df.withColumn("a", struct(df("temp") as "b")).drop("temp"),
+ "a.b", // one level nesting
+ (x: Any) => Row(x)),
+ (
+ df.withColumn("a", struct(struct(df("temp") as "c") as
"b")).drop("temp"),
+ "a.b.c", // two level nesting
+ (x: Any) => Row(Row(x))
+ ),
+ (
+ df.withColumnRenamed("temp", "a.b"),
+ "`a.b`", // zero nesting with column name containing `dots`
+ (x: Any) => x
+ ),
+ (
+ df.withColumn("a.b", struct(df("temp") as "c.d") ).drop("temp"),
+ "`a.b`.`c.d`", // one level nesting with column names containing `dots`
+ (x: Any) => Row(x)
+ )
+ ).foreach { case (df, pushDownColName, resultTransFun) =>
+ testCases(df, pushDownColName, resultTransFun)
+ }
+ }
+
private def testTimestampPushdown(data: Seq[Timestamp]): Unit = {
assert(data.size === 4)
val ts1 = data.head
val ts2 = data(1)
val ts3 = data(2)
val ts4 = data(3)
- withParquetDataFrame(data.map(i => Tuple1(i))) { implicit df =>
- checkFilterPredicate('_1.isNull, classOf[Eq[_]], Seq.empty[Row])
- checkFilterPredicate('_1.isNotNull, classOf[NotEq[_]], data.map(i =>
Row.apply(i)))
-
- checkFilterPredicate('_1 === ts1, classOf[Eq[_]], ts1)
- checkFilterPredicate('_1 <=> ts1, classOf[Eq[_]], ts1)
- checkFilterPredicate('_1 =!= ts1, classOf[NotEq[_]],
- Seq(ts2, ts3, ts4).map(i => Row.apply(i)))
-
- checkFilterPredicate('_1 < ts2, classOf[Lt[_]], ts1)
- checkFilterPredicate('_1 > ts1, classOf[Gt[_]], Seq(ts2, ts3, ts4).map(i
=> Row.apply(i)))
- checkFilterPredicate('_1 <= ts1, classOf[LtEq[_]], ts1)
- checkFilterPredicate('_1 >= ts4, classOf[GtEq[_]], ts4)
-
- checkFilterPredicate(Literal(ts1) === '_1, classOf[Eq[_]], ts1)
- checkFilterPredicate(Literal(ts1) <=> '_1, classOf[Eq[_]], ts1)
- checkFilterPredicate(Literal(ts2) > '_1, classOf[Lt[_]], ts1)
- checkFilterPredicate(Literal(ts3) < '_1, classOf[Gt[_]], ts4)
- checkFilterPredicate(Literal(ts1) >= '_1, classOf[LtEq[_]], ts1)
- checkFilterPredicate(Literal(ts4) <= '_1, classOf[GtEq[_]], ts4)
-
- checkFilterPredicate(!('_1 < ts4), classOf[GtEq[_]], ts4)
- checkFilterPredicate('_1 < ts2 || '_1 > ts3, classOf[Operators.Or],
Seq(Row(ts1), Row(ts4)))
- }
- }
-
- private def testDecimalPushDown(data: DataFrame)(f: DataFrame => Unit): Unit
= {
- withTempPath { file =>
- data.write.parquet(file.getCanonicalPath)
- readParquetFile(file.toString)(f)
+ import testImplicits._
+ withNestedDataFrame(data.toDF()) { case (df, pushDownColName, resultFun) =>
+ withParquetDataFrame(df) { implicit df =>
Review comment:
can we avoid name conflicts? there are two `df`s here. Shall we rename the
first `df` to `inputDf`?
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
