如何将“dask_cudf”列转换为日期时间？

Question

我们如何将字符串或纳秒的 dask_cudf 列转换为日期时间对象？

to_datetime

在 pandas 和 cudf 中可用。请参阅下面的示例数据

import pandas
import cudf

# with pandas
df = pandas.DataFrame( {'city' : ['Dallas','Bogota','Chicago','Juarez'], 
                      'timestamp' : [1664828099973725440,1664828099972763136,1664828094775313920,1664828081313273856]})

df['datetime'] = pd.to_datetime(df['timestamp'])

# with cdf
cdf = cudf.DataFrame( {'city' : ['Dallas','Bogota','Chicago','Juarez'], 
                      'timestamp' : [1664828099973725440,1664828099972763136,1664828094775313920,1664828081313273856]})
cdf['datetime'] = cudf.to_datetime(cdf['timestamp'])

print(df)
print(cdf) 

无论哪种情况，结果都是一样的：

      city            timestamp                      datetime
0   Dallas  1664828099973725440 2022-10-03 20:14:59.973725440
1   Bogota  1664828099972763136 2022-10-03 20:14:59.972763136
2  Chicago  1664828094775313920 2022-10-03 20:14:54.775313920
3   Juarez  1664828081313273856 2022-10-03 20:14:41.313273856

This recent SO question 建议使用 dask：

import dask_cudf
from dask import dataframe as dd

ddf = dask_cudf.from_cudf(cdf, npartitions=2)

dd.to_datetime(ddf['timestamp']).head()

产生错误。我正在从一个目录中的大量 csv 文件创建一个 dask_cudf。