Fix tests/mypy error for "not WIN" condition

[krassowski]

Attempts to address https://github.com/krassowski/jupyterlab-lsp/pull/301#issuecomment-665735786, which can be also seen in #304

[krassowski]

Well, the implications are that we were NOT unit testing paths on Windows before...

[krassowski]

We have the following errors:

>       assert normalized_uri(root_dir) == expected_root_uri
E       AssertionError: assert 'file://vboxsvr/shared-folder/' == 'file://vboxsvr/shared-folder'
E         - file://vboxsvr/shared-folder
E         + file://vboxsvr/shared-folder/
E         ?                             +

>       assert file_uri_to_path(file_uri) == expected_windows_path
E       AssertionError: assert 'C:/Windows/System32/Drivers/etc' == 'C:\\Windows\\System32\\Drivers\\etc'
E         - C:\Windows\System32\Drivers\etc
E         ?   ^       ^        ^       ^
E         + C:/Windows/System32/Drivers/etc
E         ?   ^       ^        ^       ^

>       assert file_uri_to_path(file_uri) == expected_windows_path
E       AssertionError: assert 'C:/some dir/some file.txt' == 'C:\\some dir\\some file.txt'
E         - C:\some dir\some file.txt
E         ?   ^        ^
E         + C:/some dir/some file.txt
E         ?   ^        ^

Not quite sure if it's our expectations that are incorrect, or if the implementation is wrong.

[bollwyvl]

Indeed... i need to take a look back through what prompted us to have to take these measures. I look forward to next month when we can drop 3.5, and use pathlib for many of these things.

[bollwyvl]

Sorry I didn't get to responding to this until now!

I think if the implementation was wrong, we'd be seeing more failures downstream (e.g. Foreign extractors.ipynb). I guess we could also force CI to do stuff on a different drive (there's a D: and C: on azure), and muck up the path with more strings.

Given the intent of normalized_uri, it's safe to just add the missing / for test_normalize_windows_path_case. Everywhere we use URIs, now, I'm pretty sure we're using a proper URI manipulation function, so whatever it's doing is right.

Turns out the slashes in test_file_uri_to_path_windows are also correct, as we're not doing anything path-related with them, just URL stuff, and so / should be expected. So those slashes can just be reversed in the expected, and the r taken off the front.

So yeah: just bad expectations, and boo on me for trying to be cute with ~.

[bollwyvl]

Hooray!