Minor Description
It seems inconsistent that Window...orderBy() does not accept an ascending parameter, when DataFrame.orderBy() does.
It's also slightly inconvenient since to specify a descending sort order you have to build a column object, whereas with the ascending parameter you don't.
For example:
from pyspark.sql.functions import row_number df.select( row_number() .over( Window .partitionBy(...) .orderBy('timestamp', ascending=False)))
vs.
from pyspark.sql.functions import row_number, col df.select( row_number() .over( Window .partitionBy(...) .orderBy(col('timestamp').desc())))
It would be better if Window...orderBy() supported an ascending parameter just like DataFrame.orderBy().
