如何从dask数组创建dask系列(dd.from_dask_array)

Question

我在从dask数组创建dask系列时遇到了问题：

import dask.array as da
import dask.dataframe as dd

_dict = {'doc_faturamento': ['546102424238','946102424238','777702424238'],'data_vencimento':[20190307,20190310,20190311], 'data_pagamento': [20190227,20190324,22220202],'periodo_atraso': [-8,14, 74107], 'periodo_atraso': ['PA/PD', '8-14 días', 'INAD']}

_df = pd.DataFrame( data=_dict)

_df = dd.from_pandas(_df, npartitions=2)

_peri = da.where(_df['data_pagamento']=='2222-02-02','INAD',_df['periodo_atraso'])
_peri_df = dd.from_dask_array(_peri)
_df['periodo_atraso'] = _peri

即使用这个例子，我也得到了正确的结果：

_test = da.from_array(np.arange(100000, 190000), chunks=1000)
_test_df = dd.from_dask_array(_test)

感谢您的帮助！

Answer 1

看起来像是在dask数据帧上调用da.where。我建议使用dd.DataFrame.where或dd.Series.where。