@cool-RR Thanks for this change! Am I right in thinking that this chaining syntax is only available in Python 3? Would six.raise_from() work in this case?

@martinwicke Do we have a timeline for accepting changes that use Python 3 features?

@mrry You're welcome, yes and yes.

Great! Would you mind switching the PR to use six.raise_from() and then I'd be happy to approve the change?

Yep, done. Just curious though, looking here it says that Python 3.5+ is required. What reason is there to avoid native Python 3 syntax like raise foo from bar?

Thanks! Yes, I was hoping @martinwicke could clarify our policy on that. Even though we are no longer building PIP packages for Python 2.7, I believe we still have a few non-PIP users that are stuck on that version of Python, which prevents us from adopting new syntax (just yet).

@mrry: I understand. I'll wait with fixing this problem in the entire project until after you've moved to Python 3 completely, so we could switch straight to raise foo from bar. (Assuming I won't be too busy then.) Once you've left Python 2 completely, feel free to send me a message here.

@cool-RR Any update on this PR, please. Thanks!

@gbaned See my last message.

@mrry Any update on this PR, please. Thanks!

@gbaned: See @cool-RR's last message. This PR is on hold until we get confirmation that TensorFlow can use Python 3–specific features.

@mrry, @cool-RR This PR is in draft, any update on this? Please. Thanks!

@gbaned I moved this PR to ready state now. As far as I'm concerned we can move forward, just need confirmation that Python 2 support is dropped in TensorFlow.

Here's a blog post I made where I explained my effort to push this Python 3 syntax in many open-source packages: https://blog.ram.rachum.com/post/621791438475296768/improving-python-exception-chaining-with

There's now a PyLint rule that checks for this syntax.

@mrry Can you please take a look on the above comment from @cool-RR. Thanks!

@cool-RR We are soooo close to being able to actually using Python3-only syntax. I'm re-running the tests and I will merge this if I can using the six as is.

@ematejska FYI this would make a good test case whether we can really break Py2.

@cool-RR Can you please check build failures. Thank you!

@gbaned I rebased and pushed. It's been 4 hours and the CI hasn't started running yet. Is this normal?

triggered it now

@cool-RR Still, build failures are appearing, can you please fix those?. Thank you!

Looks like you have a test test_propagates_exception_trace that checks things I don't understand about the exception traceback. Looks like @yifeif wrote it. Yifey, if you have a solution here, let me know. Otherwise this is getting too hairy for me.

Took a brief look, and seems like six changed the exception string, and you might need to update the test to reflect that. Here is the current error from CI

AssertionError: '<string>' not found in ['session.py', 'monitored_session.py'] : The exception was raised from an unrecognized file. This unit test probably needs to be updated. Traceback:
[<FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/training/monitored_session_test.py, line 750 in test_propagates_exception_trace>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/training/monitored_session.py, line 1384 in run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/six_archive/six.py, line 703 in reraise>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/training/monitored_session.py, line 1369 in run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/framework/test_util.py, line 1689 in run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/client/session.py, line 976 in run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/client/session.py, line 1202 in _run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/client/session.py, line 1380 in _do_run>, <FrameSummary file /b/f/w/bazel-out/k8-opt/bin/tensorflow/python/monitored_session_test.runfiles/org_tensorflow/tensorflow/python/client/session.py, line 1407 in _do_call>, <FrameSummary file <string>, line 3 in raise_from>]

Here is where the test was getting the name of the file exception was raised from (where the change needs to be made),

I'm gonna bow out, this is getting too complicated for me.