Skip to content
Closed
Show file tree
Hide file tree
Changes from 8 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -257,8 +257,20 @@ private[hive] class HiveMetastoreCatalog(sparkSession: SparkSession) extends Log
}
// The inferred schema may have different field names as the table schema, we should respect
// it, but also respect the exprId in table relation output.
assert(result.output.length == relation.output.length &&
result.output.zip(relation.output).forall { case (a1, a2) => a1.dataType == a2.dataType })
if (result.output.length != relation.output.length) {
throw new HiveTableConvertException(
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shall we just throw AnalysisException? It's not in the runtime stage yet.

s"Converted table has ${result.output.length} columns, " +
s"but source Hive table has ${relation.output.length} columns. " +
s"Set ${HiveUtils.CONVERT_METASTORE_PARQUET.key} to false, " +
s"or recreate table ${relation.tableMeta.identifier} to workaround.")
}
if (!result.output.zip(relation.output).forall {
case (a1, a2) => a1.dataType.sameType(a2.dataType) }) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a breaking change. We allow inferred schema may have different field names. sameType requires same field names.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok, I see. It restricts the field names by l.name == r.name. Sorry for that, I will revert this.

throw new HiveTableConvertException(
s"Column in converted table has different data type with source Hive table's. " +
s"Set ${HiveUtils.CONVERT_METASTORE_PARQUET.key} to false, " +
s"or recreate table ${relation.tableMeta.identifier} to workaround.")
}
val newOutput = result.output.zip(relation.output).map {
case (a1, a2) => a1.withExprId(a2.exprId)
}
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.spark.sql.hive

import org.apache.spark.SparkException

private[hive] case class HiveTableConvertException(message: String)
extends SparkException(message)
Original file line number Diff line number Diff line change
Expand Up @@ -358,4 +358,24 @@ class DataSourceWithHiveMetastoreCatalogSuite
Seq(table("src").count().toString))
}
}

test("SPARK-29869: Fix convertToLogicalRelation throws unclear AssertionError") {
withTempPath(dir => {
val baseDir = s"${dir.getCanonicalFile.toURI.toString}/non_partition_table"
val partitionLikeDir = s"$baseDir/dt=20191113"
spark.range(3).selectExpr("id").write.parquet(partitionLikeDir)
withTable("non_partition_table") {
withSQLConf(HiveUtils.CONVERT_METASTORE_PARQUET.key -> "true") {
spark.sql(
s"""
|CREATE TABLE non_partition_table (id bigint)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn't it a malformed table? Does hive ignore the directories for non-partitioned tables?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems Hive return none when query this table(1.2.1):

hive> select * from xxxxx.xxxx;
OK
Time taken: 25.301 seconds

But no assertion error

Copy link
Contributor Author

@LantaoJin LantaoJin Nov 14, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe you are right. Actually the table LOCATION is /path/tablename/dt=yyyymmdd, But its data file paths are /path/tablename/dt=yyyymmdd/dt=yyyymmdd/xxx.parquet. I guess Hive does not recursively lookup load the data. So it return empty but not error.
And I found if when enable recursively lookup by .option("recursiveFileLookup", true), the inferPartitioning will be disable. So dt=yyyymmdd won't be treated as partitionSpec.

So should I revert the code changes and only keep the assert detail information? Or throws exception instead of assertion, and catch it then rollback to do not use built-in Parquet reader to read?

|STORED AS PARQUET LOCATION '$baseDir'
|""".stripMargin)
val e = intercept[HiveTableConvertException](
spark.table("non_partition_table"), Seq()).getMessage
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: what does Seq() mean here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, typo

assert(e.contains("Converted table has 2 columns, but source Hive table has 1 columns."))
}
}
})
}
}