The underlying issue exposed in Issue #85 is due to a bug in a low-level function in the pandas library that converts timestamps of any unit into integers in nanoseconds. The fix in this PR should be considered temporary until the associated issue in pandas is fixed.
By passing a timestamp in nanoseconds to the pandas to_timedelta function, rather than a timestamp in seconds, the problematic code path in the pandas library is avoided.
Unit tests for the datetime_to_sow function were also included in this PR to catch this issue in the future.
The underlying issue exposed in Issue #85 is due to a bug in a low-level function in the pandas library that converts timestamps of any unit into integers in nanoseconds. The fix in this PR should be considered temporary until the associated issue in pandas is fixed.
By passing a timestamp in nanoseconds to the pandas
to_timedelta
function, rather than a timestamp in seconds, the problematic code path in the pandas library is avoided.Unit tests for the
datetime_to_sow
function were also included in this PR to catch this issue in the future.Closes Issue #85