pivot dataframe in pyspark

Question

I have DF test contains below columns

Type  Name  Country      Year    Value
1     Rec      US        2018      8
2     fg       UK        2019      2
5     vd      India      2020      1
7     se       US        2021      3

I wanted to make pivot on it I have tried below expression pivotdata=spark.sql("select * from test").groupby("Country").pivot("Year").sum("Value").show()

I am getting output But it is showing me only few columns except remaining two

Country  2018  2019  2020  2021
US        -     -
UK        -      -
India     -      -
US        -      -

So what can we do if I want all columns

In the output you posted, if you are grouping by 'country', you should have only 3 rows. why country 'us' is repeated? — kites, Jun 23 '20 at 21:32

score 2 · Accepted Answer · answered Jun 24 '20 at 06:01

If I am correctly understanding what you need, you have to provide your other columns also into the sum(). Consider the example below:

tst=sqlContext.createDataFrame([('2020-04-23',1,2,"india"),('2020-04-24',1,3,"india"),('2020-04-23',1,4,"china"),('2020-04-24',1,5,"china"),('2020-04-23',1,7,"germany"),('2020-04-24',1,9,"germany")],schema=('date','quantity','value','country'))
tst.show()
+----------+--------+-----+-------+
|      date|quantity|value|country|
+----------+--------+-----+-------+
|2020-04-23|       1|    2|  india|
|2020-04-24|       1|    3|  india|
|2020-04-23|       1|    4|  china|
|2020-04-24|       1|    5|  china|
|2020-04-23|       1|    7|germany|
|2020-04-24|       1|    9|germany|
+----------+--------+-----+-------+
df_pivot=tst.groupby('country').pivot('date').sum('quantity','value').show()
df_pivot.show()
+-------+------------------------+---------------------+------------------------+---------------------+
|country|2020-04-23_sum(quantity)|2020-04-23_sum(value)|2020-04-24_sum(quantity)|2020-04-24_sum(value)|
+-------+------------------------+---------------------+------------------------+---------------------+
|germany|                       1|                    7|                       1|                    9|
|  china|                       1|                    4|                       1|                    5|
|  india|                       1|                    2|                       1|                    3|
+-------+------------------------+---------------------+------------------------+---------------------+

If you don't like the funny column names, then you can use the agg function to define your own suffix to the pivoted column name.

tst_res=tst.groupby('country').pivot('date').agg(F.sum('quantity').alias('sum_quantity'),F.sum('value').alias('sum_value'))
tst_res.show()
+-------+-----------------------+--------------------+-----------------------+--------------------+
|country|2020-04-23_sum_quantity|2020-04-23_sum_value|2020-04-24_sum_quantity|2020-04-24_sum_value|
+-------+-----------------------+--------------------+-----------------------+--------------------+
|germany|                      1|                   7|                      1|                   9|
|  china|                      1|                   4|                      1|                   5|
|  india|                      1|                   2|                      1|                   3|
+-------+-----------------------+--------------------+-----------------------+--------------------+

pivot dataframe in pyspark

1 Answers1

Linked