StringIO and pandas read_csv

Question

I'm trying to mix StringIO and BytesIO with pandas and struggling with some basic stuff. For example, I can't get "output" below to work, whereas "output2" below does work. But "output" is closer to the real world example I'm trying to do. The way in "output2" is from an old pandas example but not really a useful way for me to do it.

import io   # note for python 3 only
            # in python2 need to import StringIO

output = io.StringIO()
output.write('x,y\n')
output.write('1,2\n')

output2 = io.StringIO("""x,y
1,2
""")

They seem to be the same in terms of type and contents:

type(output) == type(output2)
Out[159]: True

output.getvalue() == output2.getvalue()
Out[160]: True

But no, not the same:

output == output2
Out[161]: False

More to the point of the problem I'm trying to solve:

pd.read_csv(output)   # ValueError: No columns to parse from file
pd.read_csv(output2)  # works fine, same as reading from a file

Answer 1

io.StringIO here is behaving just like a file -- you wrote to it, and now the file pointer is pointing at the end. When you try to read from it after that, there's nothing after the point you wrote, so: no columns to parse.

Instead, just like you would with an ordinary file, seek to the start, and then read:

>>> output = io.StringIO()
>>> output.write('x,y\n')
4
>>> output.write('1,2\n')
4
>>> output.seek(0)
0
>>> pd.read_csv(output)
   x  y
0  1  2