nrow column isn't returned by H2OFrame().group_by().count()
from python library
You can reproduce from the python script and csv.
.count() returns a frame with 2 columns, the 2nd being "nrow"
.count() returns a frame with only 1 column, "nrow" is missing
Dropping the customerID fixes the group+count expression.
This is because by default the count aggregation is done on column with index 0 (here columnID), however as it is a string column, the grouping doesn’t apply.
Suggestion: fix the count aggregator to apply to first non-string column instead. Also need to verify what happens in R.