Previously, dividing integers by integers in the processing pipeline (e.g. with apply) would perform integer division. This is true for all of column/column, column/value, and value/column.
This is not the same as the behaviour of Pandas or Polars. Worse, if the denominator was zero, this would cause a floating point exception on linux, and an integer overflow exception (#1791) on Windows.
This change makes the result a float64 in all cases. If we want the previous behaviour to be accessible again, we should implement the integer division operator // in Python.
Reference Issues/PRs
Closes #1792 Fixes #1791
What does this implement or fix?
Previously, dividing integers by integers in the processing pipeline (e.g. with
apply
) would perform integer division. This is true for all ofcolumn/column
,column/value
, andvalue/column
. This is not the same as the behaviour of Pandas or Polars. Worse, if the denominator was zero, this would cause a floating point exception on linux, and an integer overflow exception (#1791) on Windows.This change makes the result a
float64
in all cases. If we want the previous behaviour to be accessible again, we should implement the integer division operator//
in Python.