How to write a numpy array to a csv file?

Question

I want to open up a new text file and then save the numpy array to the file. I wrote this bit of code:

foo = np.array([1,2,3])
abc = open('file'+'_2', 'w')
np.savetxt(abc, foo, delimiter=",")

I get this error:

TypeError                                 Traceback (most recent call last)
<ipython-input-33-fea41927952b> in <module>()
      2 model = cool
      3 abc = open('file'+'_2', 'w')
----> 4 np.savetxt(abc, foo, delimiter=",")

/usr/local/lib/python3.4/site-packages/numpy/lib/npyio.py in savetxt(fname, X, fmt,     delimiter, newline, header, footer, comments)
   1071         else:
   1072             for row in X:
-> 1073                 fh.write(asbytes(format % tuple(row) + newline))
   1074         if len(footer) > 0:
   1075             footer = footer.replace('\n', '\n' + comments)

TypeError: must be str, not bytes

Does anyone know whats wrong?

Additionally, I found an empty file created in the terminal called file_2, but nothing is written inside it.

EDIT: I am using Python3.4

Solution

It appears you are using Python3. Therefore, open the file in binary mode (wb), not text mode (w):

import numpy as np
foo = np.array([1,2,3])
with open('file'+'_2', 'wb') as abc:
    np.savetxt(abc, foo, delimiter=",")

Also, close the filehandle, abc, to ensure everything is written to disk. You can do that by using a with-statement (as shown above).

As DSM points out, usually when you use np.savetxt you will not want to write anything else to the file, since doing so could interfere with using np.loadtxt later. So instead of using a filehandle, it may be easier to simply pass the name of the file as the first argument to np.savetxt:

import numpy as np
foo = np.array([1,2,3])
np.savetxt('file_2', foo, delimiter=",")

Comments

[DSM]

Unless you're appending to a preexisting file, I'm not sure there's much reason to pass a handle. Might as well just pass "file_2" as the first argument to savetxt, no?