Reduce mypy errors (use sys.info instead of try-fail)

[MartinThoma]

No description provided.

[alangenfeld]

Nice, thanks! Running our full CI then will merge

[MartinThoma]

I'll have a look at that tomorrow:

************* Module dagster.seven
--
  | python_modules/dagster/dagster/seven/__init__.py:14:4: E0611: No name 'tempfile' in module 'backports' (no-name-in-module)
  | python_modules/dagster/dagster/seven/__init__.py:14:4: E0401: Unable to import 'backports.tempfile' (import-error)
  | python_modules/dagster/dagster/seven/__init__.py:30:4: E0401: Unable to import 'functools32' (import-error)

(Other topic: I'm a bit confused by seeing "buildkite/dagster/black — Passed (49 seconds)" - do you use the black formatter? Why does that pass? I'm pretty sure I have seen non-black formatted code)

[alangenfeld]

do you use the black formatter? Why does that pass?

Ya you can check the Makefile in the root to see what args we use

[MartinThoma]

When I run make check_black I get 53 files would be reformatted, 740 files would be left unchanged, 7 files would fail to reformat.

Picking just one of many examples: https://github.com/dagster-io/dagster/blob/master/python_modules/dagster/dagster/seven/abc.py#L45-L62 is not black formatting because of the ' - black would use ". Hence my confusion.

[alangenfeld]

huh - which version of black are you running? We pin to 19.10b0 via python_modules/dagster/dev-requirements.txt and use -S / --skip-string-normalization which avoids coercing quotes.

[MartinThoma]

Good catch! Yes, my black version was a bit outdated. I didn't have the -S flag.

[MartinThoma]

Going back to the bug: I'm still a bit puzzled by it. If I read https://buildkite.com/dagster/dagster/builds/8064#f1449ea8-72ea-4350-a1b6-19b68d7a7726 correctly, then it runs from a docker image 968703565975.dkr.ecr.us-west-1.amazonaws.com/buildkite-integration:py3.7.4-v6 - the name indicates that Python 3.7 is executed there. But the failure (__init__.py:14:4: E0611: No name 'tempfile' in module 'backports') indicates that sys.version_info[0] <= 2 is true.

[MartinThoma]

I understand the problem now, but it might be that it is not possible to make pylint and mypy happy at the moment. python/mypy#1393 (comment) could be a blocker. Let's see, maybe somebody knows a work-around: https://stackoverflow.com/q/59295045/562769

[alangenfeld]

not possible to make pylint and mypy happy

I think we would be fine with suppression comments to disable pylint assuming it's just in a few spots.

[mgasner]

apologies, these test failures on windows are my fault -- if you rebase on master these should go away

[mgasner]

Windows failures are resolved on master so I think this is gtg.

[mgasner]

Let's rebase on master to kick the CI off again and then should be good to go when all checks are green.