read_csv, read_table in version 0.9.0 are parsing integers as double but reporting type as int64

[nisaggarwal]

For example a file containing the text below, read using:

>>> pandas.read_csv("file.log")

produces the following output:

             Numbers
0  17007000002000192
1  17007000002000192
2  17007000002000192
3  17007000002000192
4  17007000002000192
5  17007000002000192
6  17007000002000192
7  17007000002000192
8  17007000002000192
9  17007000002000194

>> numpy.spacing(17007000002000192)

is 2.0 for this range of numbers

but the type reported for the value is int64 not double/float64

file.log contains:

Numbers
17007000002000191
17007000002000191
17007000002000191
17007000002000191
17007000002000192
17007000002000192
17007000002000192
17007000002000192
17007000002000192
17007000002000194

[wesm]

This works in 0.10.1 and git master, recommend you update

In [3]: read_clipboard()
Out[3]: 
             Numbers
0  17007000002000191
1  17007000002000191
2  17007000002000191
3  17007000002000191
4  17007000002000192
5  17007000002000192
6  17007000002000192
7  17007000002000192
8  17007000002000192
9  17007000002000194

However, it's broken with the "python parser" (engine='python'). Let me look into this quickly.