[
https://issues.apache.org/jira/browse/SPARK-27555?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hui WANG updated SPARK-27555:
-----------------------------
Description:
I already seen https://issues.apache.org/jira/browse/SPARK-17620
and https://issues.apache.org/jira/browse/SPARK-18397
and I check source code of Spark for the change of set
"spark.sql.hive.covertCTAS=true" and then spark will use
"spark.sql.sources.default" which is parquet as storage format in "create table
as select" scenario.
But my case is just create table without select. When I set
hive.default.fileformat=parquet in hive-site.xml or set
spark.hadoop.hive.default.fileformat=parquet in spark-defaults.conf, after
create a table, when i check the hive table, it still use textfile fileformat.
It seems HiveSerDe gets the value of the hive.default.fileformat parameter from
SQLConf
The parameter values in SQLConf are copied from SparkContext's SparkConf at
SparkSession initialization, while the configuration parameters in
hive-site.xml are loaded into SparkContext's hadoopConfiguration parameters by
SharedState, And all the config with "spark.hadoop" conf are setted to
hadoopconfig, so the configuration does not take effect.
was:
I already seen https://issues.apache.org/jira/browse/SPARK-17620
and https://issues.apache.org/jira/browse/SPARK-18397
and I check source code of Spark for the change of set
"spark.sql.hive.covertCTAS=true" and then spark will use
"spark.sql.sources.default" which is parquet as storage format in "create table
as select" scenario.
But my case is just create table without select. When I set
hive.default.fileformat=parquet in hive-site.xml or set
spark.hadoop.hive.default.fileformat=parquet in spark-defaults.conf, after
create a table, when i check the hive table, it still use textfile fileformat.
HiveSerDe gets the value of the hive.default.fileformat parameter from SQLConf
The parameter values in SQLConf are copied from SparkContext's SparkConf at
SparkSession initialization, while the configuration parameters in
hive-site.xml are loaded into SparkContext's hadoopConfiguration parameters by
SharedState, And all the config with "spark.hadoop" conf are setted to
hadoopconfig, so the configuration does not take effect.
> cannot create table by using the hive default fileformat in both
> hive-site.xml and spark-defaults.conf
> ------------------------------------------------------------------------------------------------------
>
> Key: SPARK-27555
> URL: https://issues.apache.org/jira/browse/SPARK-27555
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 2.3.2
> Reporter: Hui WANG
> Priority: Major
>
> I already seen https://issues.apache.org/jira/browse/SPARK-17620
> and https://issues.apache.org/jira/browse/SPARK-18397
> and I check source code of Spark for the change of set
> "spark.sql.hive.covertCTAS=true" and then spark will use
> "spark.sql.sources.default" which is parquet as storage format in "create
> table as select" scenario.
> But my case is just create table without select. When I set
> hive.default.fileformat=parquet in hive-site.xml or set
> spark.hadoop.hive.default.fileformat=parquet in spark-defaults.conf, after
> create a table, when i check the hive table, it still use textfile fileformat.
>
> It seems HiveSerDe gets the value of the hive.default.fileformat parameter
> from SQLConf
> The parameter values in SQLConf are copied from SparkContext's SparkConf at
> SparkSession initialization, while the configuration parameters in
> hive-site.xml are loaded into SparkContext's hadoopConfiguration parameters
> by SharedState, And all the config with "spark.hadoop" conf are setted to
> hadoopconfig, so the configuration does not take effect.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
