test.txt is not in your current working directory. Please, make sure test.txt and your python-file/jupyter-notebook are in same directory.Thanks!

I expect that the w mode (and w+, w+b) should have similar semantics to Python open, where w means truncate the file first.

Since this isn't the case it breaks things like numpy.savez().

Workarounds are to either write something, e.g. "", to the file before reading so that the file is created, or manually create the file first as you suggest. But those are workarounds to a problem that I think should be solved in GFile.

This happens on both nightly and 2.0.

As I'm working on modularizing filesystem support, I'm assigning this to me, although it will take a while until I can get to the python side of things.

In the end, the expected behavior should be similar to Python's:

>>> with open('this_file_does_not_exist_at_all', 'w+') as f: f.read()
... 
''

Related to the fact that GFile does not truncate files the same way as Python, reading an existing file opened with w+ will return the text in the file instead of "".

Example:

import tensorflow as tf

with open('existing.txt', 'w') as f:
    f.write('txt')
    f.flush()

with tf.io.gfile.GFile('existing.txt', 'w+') as f:
    print(f.read())  # Prints txt, should print ""

I'll have to handle that too, thanks for pointing it out

f.seek() also crashes in the same way as f.read().

Any update on this? I'm running into this error trying to write to s3 with gfile and pysoundfile

@mihaimaruseac Any updates on this? This still seems to be an issue with TF 2.5 and prevent's the use of GFile together with np.savez.

issue still exists in 2.6.0 and nightly. Here's the gist. Thanks!

Problem is the modularization effort stalled since members left the team last year.

We onboarded new members recently, so we should pick up these items again. Apologies for the delays.

from @eirism in #32090 (comment):

Workarounds are to either write something, e.g. "", to the file before reading so that the file is created, or manually create the file first as you suggest. But those are workarounds to a problem that I think should be solved in GFile.

could you give an example of such a workaround? I tried the following unsuccessful:

import numpy as np
import tensorflow as tf

with tf.io.gfile.GFile("output.npz", "w") as file:
    file.write("")
    np.savez(file, content=np.array([1, 2, 3]))

Still getting

tensorflow.python.framework.errors_impl.PermissionDeniedError: File isn't open for reading

@patzm Since numpy.savez both reads and writes to the file you need to open it in w+ mode. You get the error because w is only for writing to the file, you can not read from a w file.

sadly also doesn't work for me. Neither w+ nor wb+. Could you post a minimal example maybe?
Btw, I think this is also related to #32975.

a similar approach that works for me.

io_buffer = io.BytesIO()
np.savez(io_buffer, ...)
with gfile.Open(path, "wb") as f:
  f.write(io_buffer.getvalue())

I was able to replicate the issue in tf-nightly 2.12.0-dev20221215. Please find the gist for reference. Thank you.