Set data type for specific column when using read_csv from pandas

Question

I have a large csv file (~10GB), with around 4000 columns. I know that most of data i will expect is int8, so i set:

pandas.read_csv('file.dat', sep=',', engine='c', header=None, 
                na_filter=False, dtype=np.int8, low_memory=False)

Thing is, the final column (4000th position) is int32, is there away can i tell read_csv that use int8 by default, and at column 4000th, use int 32?

Thank you

Answer 1

If you are certain of the number you could recreate the dictionary like this:

dtype = dict(zip(range(4000),['int8' for _ in range(3999)] + ['int32']))

Considering that this works:

import pandas as pd
import numpy as np
​
data = '''\
1,2,3
4,5,6'''
​
fileobj = pd.compat.StringIO(data)
df = pd.read_csv(fileobj, dtype={0:'int8',1:'int8',2:'int32'}, header=None)
​
print(df.dtypes)

Returns:

0     int8
1     int8
2    int32
dtype: object

From the docs:

dtype : Type name or dict of column -> type, default None

Data type for data or columns. Eg {'a': np.float64, 'b': np.int32} Use str or object to preserve and not interpret dtype. If converters are specified, they will be applied INSTEAD of dtype conversion.

Answer 2

Since you have no header, the column names are the integer order in which they occur, ie the first column is df[0] . To programmatically set the last column to be int32 , you can read the first line of the file to get the width of the dataframe, then construct a dictionary of the integer types you want to use with the number of the columns as the keys.

import numpy as np
import pandas as pd

with open('file.dat') as fp:
    width = len(fp.readline().strip().split(','))
    dtypes = {i: np.int8 for i in range(width)}
    # update the last column's dtype
    dtypes[width-1] = np.int32

    # reset the read position of the file pointer
    fp.seek(0)
    df = pd.read_csv(fp, sep=',', engine='c', header=None, 
                     na_filter=False, dtype=dtypes, low_memory=False)