[jira] [Commented] (SPARK-30989) TABLE.COLUMN reference doesn't work with new columns created by UDF

2020-03-20 Thread L. C. Hsieh (Jira) 


[ 
https://issues.apache.org/jira/browse/SPARK-30989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17063779#comment-17063779
 ] 

L. C. Hsieh commented on SPARK-30989:
-

The analyzed query plan of {{df2}}:
{code}
== Analyzed Logical Plan ==
x: int, y: int, z: int
Project [x#8, y#9, UDF(y#9) AS z#12]
+- SubqueryAlias cat
 +- Project [_1#3 AS x#8, _2#4 AS y#9]
 +- SerializeFromObject [knownnotnull(assertnotnull(input[0, scala.Tuple2, 
true]))._1 AS _1#3, knownnotnull(assertnotnull(input[0, scala.Tuple2, 
true]))._2 AS _2#4]
 +- ExternalRDD [obj#2]
{code}

It is obviously that {{x}} and {{y}} are defined below {{SubqueryAlias cat}}. 
The UDF column {{z}} was added after you alias the dataframe, so it cannot be 
referred as {{cat.z}}. Doesn't it make sense?

> TABLE.COLUMN reference doesn't work with new columns created by UDF
> ---
>
> Key: SPARK-30989
> URL: https://issues.apache.org/jira/browse/SPARK-30989
> Project: Spark
>  Issue Type: Bug
>  Components: SQL
>Affects Versions: 2.4.4
>Reporter: Chris Suchanek
>Priority: Major
>
> When a dataframe is created with an alias (`.as("...")`) its columns can be 
> referred as `TABLE.COLUMN` but it doesn't work for newly created columns with 
> UDF.
> {code:java}
> // code placeholder
> df1 = sc.parallelize(l).toDF("x","y").as("cat")
> val squared = udf((s: Int) => s * s)
> val df2 = df1.withColumn("z", squared(col("y")))
> df2.columns //Array[String] = Array(x, y, z)
> df2.select("cat.x") // works
> df2.select("cat.z") // Doesn't work
> // org.apache.spark.sql.AnalysisException: cannot resolve '`cat.z`' given 
> input 
> // columns: [cat.x, cat.y, z];;
> {code}
> Might be related to: https://issues.apache.org/jira/browse/SPARK-30532



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org

[jira] [Commented] (SPARK-30989) TABLE.COLUMN reference doesn't work with new columns created by UDF

2020-03-19 Thread Wenchen Fan (Jira) 


[ 
https://issues.apache.org/jira/browse/SPARK-30989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17062355#comment-17062355
 ] 

Wenchen Fan commented on SPARK-30989:
-

https://github.com/apache/spark/pull/27916 can't fix it? I don't have a strong 
opnion as there is no clear rule about how we retain the df alias after many 
transformations.

> TABLE.COLUMN reference doesn't work with new columns created by UDF
> ---
>
> Key: SPARK-30989
> URL: https://issues.apache.org/jira/browse/SPARK-30989
> Project: Spark
>  Issue Type: Bug
>  Components: SQL
>Affects Versions: 2.4.4
>Reporter: Chris Suchanek
>Priority: Major
>
> When a dataframe is created with an alias (`.as("...")`) its columns can be 
> referred as `TABLE.COLUMN` but it doesn't work for newly created columns with 
> UDF.
> {code:java}
> // code placeholder
> df1 = sc.parallelize(l).toDF("x","y").as("cat")
> val squared = udf((s: Int) => s * s)
> val df2 = df1.withColumn("z", squared(col("y")))
> df2.columns //Array[String] = Array(x, y, z)
> df2.select("cat.x") // works
> df2.select("cat.z") // Doesn't work
> // org.apache.spark.sql.AnalysisException: cannot resolve '`cat.z`' given 
> input 
> // columns: [cat.x, cat.y, z];;
> {code}
> Might be related to: https://issues.apache.org/jira/browse/SPARK-30532



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org

[jira] [Commented] (SPARK-30989) TABLE.COLUMN reference doesn't work with new columns created by UDF

2020-03-19 Thread hemanth meka (Jira) 


[ 
https://issues.apache.org/jira/browse/SPARK-30989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17062332#comment-17062332
 ] 

hemanth meka commented on SPARK-30989:
--

[~cloud_fan] or [~viirya], can you confirm if this needs a fix? I can work on 
this if needed.

> TABLE.COLUMN reference doesn't work with new columns created by UDF
> ---
>
> Key: SPARK-30989
> URL: https://issues.apache.org/jira/browse/SPARK-30989
> Project: Spark
>  Issue Type: Bug
>  Components: SQL
>Affects Versions: 2.4.4
>Reporter: Chris Suchanek
>Priority: Major
>
> When a dataframe is created with an alias (`.as("...")`) its columns can be 
> referred as `TABLE.COLUMN` but it doesn't work for newly created columns with 
> UDF.
> {code:java}
> // code placeholder
> df1 = sc.parallelize(l).toDF("x","y").as("cat")
> val squared = udf((s: Int) => s * s)
> val df2 = df1.withColumn("z", squared(col("y")))
> df2.columns //Array[String] = Array(x, y, z)
> df2.select("cat.x") // works
> df2.select("cat.z") // Doesn't work
> // org.apache.spark.sql.AnalysisException: cannot resolve '`cat.z`' given 
> input 
> // columns: [cat.x, cat.y, z];;
> {code}
> Might be related to: https://issues.apache.org/jira/browse/SPARK-30532



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org

[jira] [Commented] (SPARK-30989) TABLE.COLUMN reference doesn't work with new columns created by UDF

2020-03-11 Thread hemanth meka (Jira) 


[ 
https://issues.apache.org/jira/browse/SPARK-30989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17057255#comment-17057255
 ] 

hemanth meka commented on SPARK-30989:
--

The alias "cat" is defined as a dataframe having 2 columns "x" and "y". The 
column "z" is generated from "cat" into a new dataframe "df2" but below code 
works and hence this exception looks like it should be the expected behaviour. 
is it not?
df2.select("z")
 

> TABLE.COLUMN reference doesn't work with new columns created by UDF
> ---
>
> Key: SPARK-30989
> URL: https://issues.apache.org/jira/browse/SPARK-30989
> Project: Spark
>  Issue Type: Bug
>  Components: SQL
>Affects Versions: 2.4.4
>Reporter: Chris Suchanek
>Priority: Major
>
> When a dataframe is created with an alias (`.as("...")`) its columns can be 
> referred as `TABLE.COLUMN` but it doesn't work for newly created columns with 
> UDF.
> {code:java}
> // code placeholder
> df1 = sc.parallelize(l).toDF("x","y").as("cat")
> val squared = udf((s: Int) => s * s)
> val df2 = df1.withColumn("z", squared(col("y")))
> df2.columns //Array[String] = Array(x, y, z)
> df2.select("cat.x") // works
> df2.select("cat.z") // Doesn't work
> // org.apache.spark.sql.AnalysisException: cannot resolve '`cat.z`' given 
> input 
> // columns: [cat.x, cat.y, z];;
> {code}
> Might be related to: https://issues.apache.org/jira/browse/SPARK-30532



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org