[
https://issues.apache.org/jira/browse/SPARK-6513?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eran Medan updated SPARK-6513:
------------------------------
Description:
It will be nice if we could treat a Dataframe just like an RDD (wherever it
makes sense)
*Worked in 1.2.1*
{code}
val sqlContext = new HiveContext(sc)
import sqlContext._
val jsonRDD = sqlContext.jsonFile(jsonFilePath)
jsonRDD.registerTempTable("jsonTable")
val jsonResult = sql(s"select * from jsonTable")
val foo = jsonResult.zipWithUniqueId().map {
case (Row(...), uniqueId) => // do something useful
...
}
foo.registerTempTable("...")
{code}
*Stopped working in 1.3.0*
{code}
jsonResult.zipWithUniqueId() //since RDDApi doesn't implement that method
{code}
**Not working workaround:**
although this might give me an {{RDD\[Row\]}}:
{code}
jsonResult.map(identity).zipWithUniqueId()
{code}
Now this won't work obviously since {{RDD\[Row\]}} does not have a
{{registerTempTable}} method of course
{code}
foo.registerTempTable("...")
{code}
(see related SO question:
http://stackoverflow.com/questions/29243186/is-this-a-regression-bug-in-spark-1-3)
EDIT: changed from issue to enhancement request
was:
I'm sure this has an Issue somewhere but I can't find it.
I see this is not a regression bug (since Ap, since it compiled in 1.2.1 but
stopped in 1.3 without any earlier deprecation warnings, but I am sure the
authors are well aware, so please change it to an enhancement request if you
disagree this is a regression. It's such an obvious and blunt regression that I
doubt it was done without a lot of thought and I'm sure there was a good
reason, but still it breaks my code and I don't have a workaround :)
Here are the details / steps to reproduce
*Worked in 1.2.1* (without any deprecation warnings)
{code}
val sqlContext = new HiveContext(sc)
import sqlContext._
val jsonRDD = sqlContext.jsonFile(jsonFilePath)
jsonRDD.registerTempTable("jsonTable")
val jsonResult = sql(s"select * from jsonTable")
val foo = jsonResult.zipWithUniqueId().map {
case (Row(...), uniqueId) => // do something useful
...
}
foo.registerTempTable("...")
{code}
*Stopped working in 1.3.0* (simply does not compile, and all I did was change
to 1.3)
{code}
jsonResult.zipWithUniqueId() //since RDDApi doesn't implement that method
{code}
**Not working workaround:**
although this might give me an {{RDD\[Row\]}}:
{code}
jsonResult.map(identity).zipWithUniqueId()
{code}
Now this won't work obviously since {{RDD\[Row\]}} does not have a
{{registerTempTable}} method of course
{code}
foo.registerTempTable("...")
{code}
(see related SO question:
http://stackoverflow.com/questions/29243186/is-this-a-regression-bug-in-spark-1-3)
> Add zipWithUniqueId (and other RDD APIs) to RDDApi.scala
> --------------------------------------------------------
>
> Key: SPARK-6513
> URL: https://issues.apache.org/jira/browse/SPARK-6513
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 1.3.0
> Environment: Windows 7 64bit, Scala 2.11.6, JDK 1.7.0_21 (though I
> don't think it's relevant)
> Reporter: Eran Medan
> Priority: Minor
>
> It will be nice if we could treat a Dataframe just like an RDD (wherever it
> makes sense)
> *Worked in 1.2.1*
> {code}
> val sqlContext = new HiveContext(sc)
> import sqlContext._
> val jsonRDD = sqlContext.jsonFile(jsonFilePath)
> jsonRDD.registerTempTable("jsonTable")
> val jsonResult = sql(s"select * from jsonTable")
> val foo = jsonResult.zipWithUniqueId().map {
> case (Row(...), uniqueId) => // do something useful
> ...
> }
> foo.registerTempTable("...")
> {code}
> *Stopped working in 1.3.0*
> {code}
> jsonResult.zipWithUniqueId() //since RDDApi doesn't implement that method
> {code}
> **Not working workaround:**
> although this might give me an {{RDD\[Row\]}}:
> {code}
> jsonResult.map(identity).zipWithUniqueId()
> {code}
> Now this won't work obviously since {{RDD\[Row\]}} does not have a
> {{registerTempTable}} method of course
> {code}
> foo.registerTempTable("...")
> {code}
> (see related SO question:
> http://stackoverflow.com/questions/29243186/is-this-a-regression-bug-in-spark-1-3)
> EDIT: changed from issue to enhancement request
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
