Column alias after groupBy in pyspark

Question

I need the resulting data frame in the line below, to have an alias name "maxDiff" for the max('diff') column after groupBy. However, the below line does not makeany change, nor throw an error.

 grpdf = joined_df.groupBy(temp1.datestamp).max('diff').alias("maxDiff")

Solution 1

You can use agg instead of calling max method:

from pyspark.sql.functions import max

joined_df.groupBy(temp1.datestamp).agg(max("diff").alias("maxDiff"))

Similarly in Scala

import org.apache.spark.sql.functions.max

joined_df.groupBy($"datestamp").agg(max("diff").alias("maxDiff"))

or

joined_df.groupBy($"datestamp").agg(max("diff").as("maxDiff"))

Solution 2

This is because you are aliasing the whole DataFrame object, not Column. Here's an example how to alias the Column only:

import pyspark.sql.functions as func

grpdf = joined_df \
    .groupBy(temp1.datestamp) \
    .max('diff') \
    .select(func.col("max(diff)").alias("maxDiff"))

Solution 3

In addition to the answers already here, the following are also convenient ways if you know the name of the aggregated column, where you don't have to import from pyspark.sql.functions:

1

grouped_df = joined_df.groupBy(temp1.datestamp) \
                      .max('diff') \
                      .selectExpr('max(diff) AS maxDiff')

See docs for info on .selectExpr()

2

grouped_df = joined_df.groupBy(temp1.datestamp) \
                      .max('diff') \
                      .withColumnRenamed('max(diff)', 'maxDiff')

See docs for info on .withColumnRenamed()

This answer here goes into more detail: https://stackoverflow.com/a/34077809

Solution 4

you can use.

grouped_df = grpdf.select(col("max(diff)") as "maxdiff",col("sum(DIFF)") as "sumdiff").show()

View more solutions

Comments

[mhn]

yes.. this works.. except that func. is not required

[Nhor]

it's not if you have already imported pyspark.sql.functions to your script

[abeboparebop]

I prefer this to the accepted answer because it doesn't require one to know in advance the name that Spark will give to the aggregated column.

[serkan kucukbay]

Did you try first code block? It seems wrong. Spark can not resolve column name with mentioned ways. In your code block, spark try to find diff column and try to run max function on given set but grouped_data doesn't contain any diff column, it contains temp1.datestamp and max(diff).

[user1978816]

You do have to be careful, as the following syntax does not work: rmse = df.groupBy('name').agg({'sqerr':'mean'}).alias('MSE') The alias command just seems to be ignored. But this works: df.groupBy('name').agg(F.mean(df.sqerr).alias('MSE'))

[user1978816]

The problem is you need to know what spark will clal the new column which might not be obvious. @zero323's anwser is better