从HTTP标头的每一列中指定read_csv中每一列的数据类型

Question

project_id = request.data['project']
list_fields = request.POST.getlist('headers')
type_fields = request.POST.getlist('type')

dataframe = pandas.read_csv(file_path, header=0)
                for field in list_fields:
                    for tipo in type_fields:
                        dataframe[field] = dataframe[field].astype(type)

how can I assign each type of data to a column according to the past in the request?

Answer 1

You can pass a list from the front with all the columns that you want to define, than pass another list with all the types that you want for the columns. After that you can cast those loops.

to_define_list_fields = request.POST.getlist('define')

type_list_fields = request.POST.getlist('types')                  

for dfield in to_define_list_fields:

   for type in type_list_fields:

       dataframe[dfield] = dataframe[dfield].astype(type)

Answer 2

When you use from_csv() , pandas will do an awful lot of type inference. In fact, more so than other methods such as convert_objects . I asked a question about it here , which is somewhat related.

I'm presuming that the front-end user has to specify the dtype of every column. In which case, it's a simple case of:

import numpy as np
import pandas as pd

df = pd.DataFrame({'a':[1, 2, 3], 'b': [4, 5, 6], 'c': [7, 8, 9]}, 
                   dtype=int)

list_fields = ['a', 'b', 'c']
list_types = [str, int, np.float64]

for field, dtype in zip(list_fields, list_types):
    df[field] = df[field].astype(dtype)

print(df.dtypes)

If the users don't have to specify the dtype of all of the fields then, on further thought, I think that would be a different topic entirely on how you filter/process user input.