pyspark Column is not iterable

Question

Having this dataframe I am getting Column is not iterable when I try to groupBy and getting max:

linesWithSparkDF
+---+-----+
| id|cycle|
+---+-----+
| 31|   26|
| 31|   28|
| 31|   29|
| 31|   97|
| 31|   98|
| 31|  100|
| 31|  101|
| 31|  111|
| 31|  112|
| 31|  113|
+---+-----+
only showing top 10 rows


ipython-input-41-373452512490> in runlgmodel2(model, data)
     65     linesWithSparkDF.show(10)
     66 
---> 67     linesWithSparkGDF = linesWithSparkDF.groupBy(col("id")).agg(max(col("cycle")))
     68     print "linesWithSparkGDF"
     69 

/usr/hdp/current/spark-client/python/pyspark/sql/column.py in __iter__(self)
    241 
    242     def __iter__(self):
--> 243         raise TypeError("Column is not iterable")
    244 
    245     # string methods

TypeError: Column is not iterable

Answer 1

It's because, you've overwritten the max definition provided by apache-spark, it was easy to spot because max was expecting an iterable.

To fix this, you can use a different syntax, and it should work.

inesWithSparkGDF = linesWithSparkDF.groupBy(col("id")).agg({"cycle": "max"})

or alternatively

from pyspark.sql.functions import max as sparkMax

linesWithSparkGDF = linesWithSparkDF.groupBy(col("id")).agg(sparkMax(col("cycle")))

Answer 2

The idiomatic style for avoiding this problem -- which are unfortunate namespace collisions between some Spark SQL function names and Python built-in function names -- is to import the Spark SQL functions module like this:

from pyspark.sql import functions as F 
# USAGE: F.col(), F.max(), F.someFunc(), ...

Then, using the OP's example, you'd simply apply F like this:

linesWithSparkGDF = linesWithSparkDF.groupBy(F.col("id")) \
                               .agg(F.max(F.col("cycle")))

In practice, this is how the problem is avoided idiomatically. =:)

Answer 3

I know the question is old but this might help someone.

First import the following :

from pyspark.sql import functions as F

Then

linesWithSparkGDF = linesWithSparkDF.groupBy(col("id")).agg(F.max(col("cycle")))