package tango

[h-vetinari]

Gonna be necessary for the allennlp-stack, see e.g. here

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipes/tango) and found it was in an excellent condition.

[h-vetinari]

@dirkgr @epwalsh could you have a look here?
I'm already skipping TestMetadata (because checking whether allenai/tango is a remote really makes no sense outside of your development setup), but I cannot determine whether the following failure (OSX-only) is relevant:

____________________ TestMain.test_deterministic_experiment ____________________
[...]
        # Check metadata.
        metadata_path = self.TEST_DIR / "hello_world" / "executor-metadata.json"
        assert metadata_path.is_file()
        metadata_params = Params.from_file(metadata_path)
        metadata = ExecutorMetadata.from_params(metadata_params)
        assert metadata.config == {
            "type": "concat_strings",
            "string1": {"type": "ref", "ref": "hello"},
            "string2": "World!",
            "join_with": ", ",
        }
>       assert metadata.git.commit is not None
E       AttributeError: 'NoneType' object has no attribute 'commit'

Seems that ExecutorMetadata does not create a member metadata.git, but I'm not sure if that's relevant?

[dirkgr]

I think it's possible that this test can't run outside of a git directory. I'm taking a look.

[h-vetinari]

I think it's possible that this test can't run outside of a git directory.

I have git running in the test environment, and the test passes on linux, but not on osx for some reasons.

I'm taking a look.

Thanks!

[dirkgr]

Skipping TestMetadata makes sense, but it would be nice if we could keep test_deterministic_experiment. Is there an environment variable we could look for that's set in conda builds?

[dirkgr]

Actually, I can check for the presence of the .git directory. If it's there, those members should be filled.

[dirkgr]

@h-vetinari, how about allenai/tango#99?

[h-vetinari]

@h-vetinari, how about allenai/tango#99?

Fine by me! :)

[h-vetinari]

@conda-forge/staged-recipes
This one should be ready.

[xhochy]

Should this have a pip check somewhere?

[h-vetinari]

Should this have a pip check somewhere?

Let me ask differently - given a passing test suite, what useful information does pip check still cover? I get that it's used often where the upstream test suite is not run, but would you consider it necessary (or even just best practice) if the test suite is being run already?

(not against adding a pip check, just trying to clarify)

[xhochy]

It checks "roughly" for the correct pinnings. In some cases, tests still pass although the pinnings conflict. In some cases (Tensorflow :D ) the upstream pinnings are much tighter than what works and thus we forget them in the recipe. This is not a problem for the package itself but for packages consuming it.

[xhochy]

"but for packages consuming it." … and have a pip check but no tests.