Python fromfile returning incorrect values when reading binary file generated with tofile

I found several similar questions but none of them described exactly the problem I'm facing. I've an OBJ file which contains lines describing vertex positions:

v 0.01214 0.4242 0.82874

I want to transform this data into a matlab-friendly format so that I can plot some data with these and other values obtained from other files. I'm using python (and numpy) to transform this data to a binary file so I use the following function:

def extract_verts(in_file, out_file):
    v = np.zeros(3, np.float32)
    for line in in_file:
        words = line.split()
        if words[0] == 'v':
           v[:] = [np.float32(s) for s in words[1:4]]
           print v[:]
           v.tofile(out_file)

The problem is that when I read it using matlab (fread does the job) the first values are correct but then incorrect values are read. after some incorrect values correct values are read again, but they see开发者_开发问答m to be shifted (when using a vector-3 structure the x-component appears as the y-component, for example). Later, it happens again and incorrect and then correct but shifted values are read, and so on. I've checked that the file is read correctly since I can see the read values from the print line.

I've tried to read the file with python, just in case it was a matlab issue:

data = np.fromfile('file.dat', np.float32)
i = 0
while i < 100:
     print data[i]
     i = i+1

and it happens exactly the same (even the incorrect values are the same).

I think it may have something to do with byte ordering or OS dependent issues, because this set of scripts I'm using works on MacOS (the scripts were created by a colleague), but I'm using Windows 7. Has anyone faced a similar problem in the past?

Thanks.

---

Most probably, the error is in the code you din't show. The out_file parameter to extract_verts() seems to be an open file object, and it is important that you open this file in binary mode:

out_file = open("out.dat", "wb")

Files are opened in text mode by default. On Windows, this means automatic newline mangling takes place: All occurences of the byte '\n' are replaced by '\r\n'. This is why all values up to the first occurrence of '\n' are correct.