How to get pandas.read_csv not to perform any conversions?

Question

For example, the values in '/tmp/test.csv' (namely, 01, 02, 03) are meant to represent strings that happen to match /^\d+$/, as opposed to integers:

In [10]: print open('/tmp/test.csv').read()
A,B,C
01,02,03

By default, pandas.read_csv converts these values to integers:

In [11]: import pandas

In [12]: pandas.read_csv('/tmp/test.csv')
Out[12]: 
   A  B  C
0  1  2  3

I want to tell pandas.read_csv to leave all these values alone. I.e., perform no conversions whatsoever. Furthermore, I want this "please do nothing" directive to be applied across-the-board, without my having to specify any column names or numbers.

I tried this, which achieved nothing:

In [13]: import csv

In [14]: pandas.read_csv('/tmp/test.csv', quoting=csv.QUOTE_ALL)
Out[14]: 
   A  B  C
0  1  2  3

The only thing that worked was to define a big ol' ConstantDict class, and use an instance of it that always returns the identity function (lambda x: x) as the value for the converters parameter, and thereby trick pandas.read_csv into doing nothing:

In [15]: %cpaste
class ConstantDict(dict):
    def __init__(self, value):
        self.__value = value
    def get(self, *args):
        return self.__value
--
Pasting code; enter '--' alone on the line to stop or use Ctrl-D.
::::::
In [16]: pandas.read_csv('/tmp/test.csv', converters=ConstantDict(lambda x: x))
Out[16]: 
    A   B   C
0  01  02  03

That's a lot of gymnastics to get such a simple "please do nothing" request across. (It would be even more gymnastics if I were to make ConstantDict bullet-proof.)

Isn't there a simpler way to achieve this?

Answer 1

df = pd.read_csv('temp.csv', dtype=str)

From the docs:

dtype : Type name or dict of column -> type, default None
Data type for data or columns. E.g. {‘a’: np.float64, ‘b’: np.int32} (Unsupported with engine=’python’). Use str or object to preserve and not interpret dtype.