I would like to load the result of numpy.savetxt into a string. Essentially the following code without the intermediate file:
import numpy as np
def savetxts(arr):
np.savetxt('tmp', arr)
with open('tmp', 'rb') as f:
return f.read()
Add a comment
|
4 Answers
For Python 3.x you can use the io module:
>>> import io
>>> s = io.BytesIO()
>>> np.savetxt(s, (1, 2, 3), '%.4f')
>>> s.getvalue()
b'1.0000\n2.0000\n3.0000\n'
>>> s.getvalue().decode()
'1.0000\n2.0000\n3.0000\n'
Note: I couldn't get io.StringIO() to work. Any ideas?
Comments
You can use StringIO (or cStringIO):
This module implements a file-like class, StringIO, that reads and writes a string buffer (also known as memory files).
The description of the module says it all. Just pass an instance of StringIO to np.savetxt instead of a filename:
>>> s = StringIO.StringIO()
>>> np.savetxt(s, (1,2,3))
>>> s.getvalue()
'1.000000000000000000e+00\n2.000000000000000000e+00\n3.000000000000000000e+00\n'
>>>
Have a look at array_str or array_repr: http://docs.scipy.org/doc/numpy/reference/routines.io.html
1 Comment
Tom Pohl
By default, both functions shorten their output which is typically not what you want. You can tweak their output using
numpy.set_printoptions, but that feels like a hack.Just requires extending previous answers with decode to UTF8 in order to generate a string. Very useful for exporting data to human readable text files.
import io
import numpy as np
s = io.BytesIO()
np.savetxt(s, np.linspace(0,10, 30).reshape(-1,3), delim=',' '%.4f')
outStr = s.getvalue().decode('UTF-8')
