search

chendaniely / pandas_for_everyone

Repository to accompany "Pandas for Everyone"
http://a.co/d/c270uul
MIT License
389 stars 402 forks source link

timedelta64[Y] on p.49 is no longer supported in pandas 2? What alternatives are available? #20

Open ka2bashi opened 10 months ago

ka2bashi commented 10 months ago

The following code appears on p.49 in 2nd edition.

scientists['age_years'] = (scientists['age_days'].astype('timedelta64[Y]'))

When this is executed in pandas 2.0.3(Python 3.11.5), the following error is output.

Cannot convert from timedelta64[ns] to timedelta64[Y]. Supported resolutions are 's', 'ms', 'us', 'ns

It seems that timedelta64[Y] is no longer supported in pandas 2.0.0 released in April. This seems to be the cause of the error. https://pandas.pydata.org/docs/dev/whatsnew/v2.0.0.html#construction-with-datetime64-or-timedelta64-dtype-with-unsupported-resolution

Can the following code be considered as an alternative to the above code?

# Calculate a year in seconds
seconds_in_a_year = 60 * 60 * 24 * 365

# convert timedelta to seconds and then to years
scientists['age_years'] = scientists['age_days'].dt.total_seconds() / seconds_in_a_year

print(scientists)

Please let us know if there is a better way. Thanks.