Use default when None given to tag_class attribute

[Silic0nS0ldier]

With this, code like the following is valid.

pip = use_extension("@rules_python//python/extensions:pip.bzl", "pip")
pip.parse(
    hub_name = "bazel_pip_dev_deps",
    python_version = "3.8",
    requirements_lock = "//:requirements.txt",
    parallel_download = None,
)

Like with rules and repository_rules, the default (True in the example) will be applied.

[meteorcloudy]

Wouldn't this be a problem is the user actually wants to pass None to the attribute?

[meteorcloudy]

Like with rules and repository_rules, the default (True in the example) will be applied.

I see, this is consistent with rules and repository_rules. Then LGTM!

[Silic0nS0ldier]

Wouldn't this be a problem is the user actually wants to pass None to the attribute?

As noted, this is for consistency. #23548 also makes None a possible input, in a round-about way (None usable as default). Allowing None to be truly passed through seems like more trouble than it'd be worth, with reasoning given on the aforementioned PR.

[fmeum]

@Silic0nS0ldier Apart from consistency, what's the use case you have in mind for this? In the example above, both removing the argument and setting it to True explicitly seem cleaner to me.

For build rules and repository rules, being able to use None to force a default value is important because it simplifies macros wrapping such rules. But that's not relevant for module extension tags, which don't support macros.

[Wyverald]

I have the same question as @fmeum. It's unclear to me what this is useful for.

[Silic0nS0ldier]

No specific use case in mind here, this was more of a knee-jerk response to noticing what is on the surface divergent behaviour for the attr module.

As it stands right now, there is no need to support None attributes for tags in MODULE.bazel.

I do think that this divergence should be documented. Perhaps in tag_class and attr docs. I'll close this and raise a PR that does just that.