[
https://issues.apache.org/jira/browse/SPARK-25948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16680810#comment-16680810
]
Hyukjin Kwon commented on SPARK-25948:
--------------------------------------
Please avoid to set {{Critical}}+ in priority which is usually reserved for
committers.
> Spark load floating point number is automatically rounded to an integer
> -----------------------------------------------------------------------
>
> Key: SPARK-25948
> URL: https://issues.apache.org/jira/browse/SPARK-25948
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 2.3.0
> Environment: redshift
> jdk 1.8.0_192 (oracle)
> scala 2.11.12
> spark 2.3.0
> Reporter: e
> Priority: Major
>
> When you use spark load redshift to aggregate data, the floating point number
> is automatically rounded to an integer.
> {code:sql}
> drop table tmp.t1 ;
> create table tmp.t1 (
> quantity integer,
> amount numeric(25,4)
> );
> insert into tmp.t1 values (1,11);
> insert into tmp.t1 values (1,11.2);
> insert into tmp.t1 values (2,2.74);
> {code}
> {code}
> val ss = SparkSession.builder()
> .appName("tmp")
> .master("local[12]")
> .getOrCreate()
> val database="REDSHIFT"
> val dconfig=GetConf()
> val sql=
> """
> |(select sum(amount*quantity ) amount,sysdate load_time from tmp.t1)
> """.stripMargin
> val sql1=
> """
> |(select sum(amount*quantity ) amount from tmp.t1)
> """.stripMargin
>
> val df = ss.read.format("jdbc").options(Map(
> "url" -> JdbcUrl(database),
> "user" -> dconfig.getString(s"$database.CONN_USER"),
> "password" -> dconfig.getString(s"$database.CONN_PASSWORD"),
> "driver" -> dconfig.getString(s"$database.CONN_DRIVER"),
> "dbtable"-> sql
> )).load().repartition(1)
> val df1 = ss.read.format("jdbc").options(Map(
> "url" -> JdbcUrl(database),
> "user" -> dconfig.getString(s"$database.CONN_USER"),
> "password" -> dconfig.getString(s"$database.CONN_PASSWORD"),
> "driver" -> dconfig.getString(s"$database.CONN_DRIVER"),
> "dbtable"-> sql1
> )).load().repartition(1)
> println(df.show(false))
> println(df1.show(false))
> result：
> +------+--------------------------+
> |amount|load_time |
> +------+--------------------------+
> |28 |2018-11-06 14:12:46.184986|
> +------+--------------------------+
> +-------+
> |amount |
> +-------+
> |27.6800|
> +-------+
> {code}
> h5. different :
> {code:sql}
> select sum(amount*quantity ) amount,sysdate load_time from tmp.t1
> select sum(amount*quantity ) amount from tmp.t1
> {code}
--
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
