1

i have done this code , my question is for the function cast data type , how can i cast all column included in a dataset at the same execept the column time stamp , and the other question is how to apply function avg on all column except also column timestamp. Thanks a lot 

val df = spark.read.option("header",true).option("inferSchema", "true").csv("C:/Users/mhattabi/Desktop/dataTest.csv")
val result=df.withColumn("new_time",((unix_timestamp(col("time")) /300).cast("long") * 300).cast("timestamp"))
result("value").cast("float")//here the first question 
val finalresult=result.groupBy("new_time").agg(avg("value")).sort("new_time")//here the second question about avg
finalresult.coalesce(1).write.format("com.databricks.spark.csv").option("header", "true").save("C:/mydata.csv")
user7394882
  • 193
  • 5
  • 17
  • Can't you just add `withColumn` for every column you want to cast? And as many `avg` as columns inside `agg`? – Mariusz Mar 01 '17 at 09:18
  • @Mariusz The proble that the dataset is very big , and there much column ,just want to do some thing that took all column except the column time – user7394882 Mar 01 '17 at 10:09

1 Answers1

0

This is quite easy to implement in pyspark, but I run into touble trying to rewrite this to scala code... I hope you will manage it somehow.

from pyspark.sql.functions import *
df = spark.createDataFrame([(100, "4.5", "5.6")], ["new_time", "col1", "col2"])
columns = [col(c).cast('float') if c != 'new_time' else col(c) for c in df.columns]
aggs = [avg(c) for c in df.columns if c != 'new_time']
finalresult = df.select(columns).groupBy('new_time').agg(*aggs)
finalresult.explain()

*HashAggregate(keys=[new_time#0L], functions=[avg(cast(col1#14 as double)), avg(cast(col2#15 as double))])
+- Exchange hashpartitioning(new_time#0L, 200)
   +- *HashAggregate(keys=[new_time#0L], functions=[partial_avg(cast(col1#14 as double)), partial_avg(cast(col2#15 as double))])
      +- *Project [new_time#0L, cast(col1#1 as float) AS col1#14, cast(col2#2 as float) AS col2#15]
         +- Scan ExistingRDD[new_time#0L,col1#1,col2#2]
Mariusz
  • 13,481
  • 3
  • 60
  • 64