Allow specifying SimpleNamespace attributes as a mapping/iterable in a positional argument

[serhiy-storchaka]

Feature or enhancement

Has this already been discussed elsewhere?

I have already discussed this feature proposal on Discourse

Links to previous discussion of this feature:

#96145
https://mail.python.org/archives/list/python-dev@python.org/message/TYIXRGOLWGV5TNJ3XMV443O3RWLEYB65/

Proposal:

It was discussed in the context of adding controversial AttrDict. The reason for adding AttrDict was to allow JSON object attributes to be accessed as Python object attributes rather than dict keys. The reason for not using SimpleNamespace for this was that you need to use a wrapper lambda x: SimpleNamespace(**x), which makes this solution less obvious.

Supporting a positional argument in SimpleNamespace(), which can be a mapping or an iterable of key-value pairs (like in the dict constructor) will make using SimpleNamespace with JSON more convenient.

SimpleNamespace({'a': 1, 'b': 2}) is the same as SimpleNamespace([('a', 1), ('b', 2)]) is the same as SimpleNamespace(a=1, b=2).

Linked PRs

• gh-108191: Add support of positional argument in SimpleNamespace constructor #108195

[serhiy-storchaka]

@rhettinger, @ericsnowcurrently, I would like to know your opinion about this feature.

Raymond, would it be useful to you, in any way?

Eric, is it compatible with your vision of SimpleNamespace?

If both answers are "yes", I'll merge the PR. If both answers are "no", I'll withdraw it and close the issue. If answers are mixed, we should have a discussion.

[ericsnowcurrently]

Yeah, I'm fine with this. It didn't seem worth it originally but it certainly isn't a bad thing.

[vstinner]

Congrats for landing this new feature.

[serhiy-storchaka]

It is a minor feature. It was hanging for such long time only because I was not sure that it is worthy.