Because VALID_MODULE_NAME is r'[_a-z]\w*\.py$' in unittest/loader.py.

Using r'[^\W\d]\w*\.py$' insteaded.

Are the module names valid in import statements?

it would help if you could perhaps attach a little tar/zip file with an example failure.

There is an attached file for examples.

I ran
{{{
cd test_dir
python -m unittest -v
}}}

and got
"Ran 1 test in 0.000s"

By the way, I ran with Python 3.4.0.

r'[^\W\d]\w*' doesn't match all valid Python identifiers. It would be more correct to write the check as:

    root, ext = os.path.splitext(basename)
    if not (ext == '.py' and root.isidentifier()):
        # valid Python identifiers only
        return None, False

Yes, I bet that regex is left over from python2, where we didn't have isidentifier.

Thank you.
I updated my patch in `VALID_MODULE_NAME.patch`.

update by adding `except AttributeError:`

Thank you very much for writing your patch in backwards compatible style - it will make backporting to unittest2 much easier.

I'm torn on whether this needs a test or not. It would be hard to regress, but testing this properly really wants hypothesis with a valid-python-identifier-strategy.

I think on balance we do need one.

So - we need a test in test_discover that mocks the presence of a file with a name containing e.g. \u2603.

sih4sing5hong5  - I think we do need a test in fact - it can be done using mocks, but right now I think the patch has a bug - it looks for isidentifier on $thing.py, but not on just $thing (which we need to do to handle packages, vs modules).

Add PR: https://github.com/python/cpython/pull/1338/


rbcollins: Need for help to review the patch, I think that both `$thing` and `$thing.py` can't be used in python (and for UNIX dir), and `\u2603` (☃) though can do something like `☃.py`, but it is not a valid identifier in python, too.

The original PR refers to a branch that no longer exists, but the behaviour documented still applies to master. There were some changes to the test loader, but none that fixed this issue.

I've opened a new PR at https://github.com/python/cpython/pull/13149 with the commit from https://github.com/python/cpython/pull/1338 and some additional changes to address the review comments given by serhiy.storchaka and rbcollins

thanks, will wait for a review from Serhiy, Rbcollins or ezio

What is the current error on test_dir.tar.gz? I'm not sure which problem is trying to be solved here.

Why does PR 13149 use str.isidentifier() method? unittest doesn't allow arbitrary Unicode in filenames?

From the description, I think the bug is that filenames that *begin* with non-ascii are not searched for tests.  Looking at the test_dir.tar.gz contents,  this is the test case that I'd use:

Broken:

$ python3 -m unittest discover -vv -p '*.py'
test_走 (tests試驗.Test試驗.試驗) ... ok
test_走 (tests試驗.test試驗.試驗) ... ok

----------------------------------------------------------------------
Ran 2 tests in 0.000s

OK

Corrected:
$ /srv/python/cpython/python -m unittest discover -vv -p '*.py'
test_走 (tests試驗.Test試驗.試驗) ... ok
test_走 (tests試驗.test試驗.試驗) ... ok
test_走 (tests試驗.試驗.試驗) ... ok

----------------------------------------------------------------------
Ran 3 tests in 0.000s

OK


isidentifier() is used because filenames to be discovered must be importable and thus valid identifiers:  https://docs.python.org/3/library/unittest.html#test-discovery