I have a user-defined function tmp_func and a dask dataframe df. I would like to apply this function on each group of df.

def tmp_func(s_df):
    ...
    return(s_df)

result = df.groupby('id').apply(tmp_func, meta = meta)
result = result.compute(scheduler = 'processes')

It is recommended to specify the dtypes of columns of dataframe returned by tmp_func. In my case, the resulting dataframe from tmp_func has over 20.000 columns which contain only natural numbers. So I think np.int8 is the datatype.

Is there anyway to specify that all columns have the same datatype np.int8? It would be a nightmare to specify it by a dictionary with over 20.000 elements.

CodePudding user response：

Just use a dict-comprehension

result = df.groupby('id').apply(tmp_func, meta = {col: np.int8 for col in df.columns)}