dataclasses: Allow typing.Annotated to wrap dataclasses-specific annotations

[GBeauregard]

BPO	46511
Nosy	@ericvsmith, @JelleZijlstra, @GBeauregard
PRs	gh-90669: Let typing.Annotated wrap dataclasses annotations #30997

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2022-01-25.09:02:48.629>
labels = ['type-bug', 'library', '3.9', '3.10', '3.11']
title = 'dataclasses: Allow typing.Annotated to wrap dataclasses-specific annotations'
updated_at = <Date 2022-02-11.19:22:33.744>
user = 'https://github.com/GBeauregard'

bugs.python.org fields:

activity = <Date 2022-02-11.19:22:33.744>
actor = 'GBeauregard'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2022-01-25.09:02:48.629>
creator = 'GBeauregard'
dependencies = []
files = []
hgrepos = []
issue_num = 46511
keywords = ['patch']
message_count = 9.0
messages = ['411567', '411614', '411637', '411709', '411710', '411945', '411948', '412033', '413094']
nosy_count = 3.0
nosy_names = ['eric.smith', 'JelleZijlstra', 'GBeauregard']
pr_nums = ['30997']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue46511'
versions = ['Python 3.9', 'Python 3.10', 'Python 3.11']

[GBeauregard]

In https://bugs.python.org/issue46491 the typing runtime behavior was changed so that Annotated[Classvar[...]] is now valid at runtime in order to alleviate tension between typing and non-typing annotation space uses. dataclasses.py should likely follow suit in its runtime use of ClassVar and InitVar.

Reviewing the code I see two areas that would need addressed:

1. InitVar needs changed so Annotated[InitVar[...]] is no longer a runtime error. This is currently a runtime error completely by accident: typing.py expects special type forms to be callable(), usually by implementing a __call__ that throws an error, but InitVar does not implement this. Adding an implementation like in typing.py would fix the runtime error:
   

   cpython/Lib/typing.py

   Lines 391 to 392 in b1a3446

   	def __call__(self, *args, **kwds):
   	raise TypeError(f"Cannot instantiate {self!r}")

2. The dataclasses-specific typehint introspection implementation needs modified to accommodate being wrapped by an Annotated annotation. I see in the comments the code is performance sensitive so I'm not sure what you want to do; f.ex. the regex needs modified, but it's not clean.

What are your thoughts?

[JelleZijlstra]

Related: bpo-44799 about InitVar.

[ericvsmith]

My thoughts are that I'd like PEP-563 to go away, and PEP-649 to be accepted, and also never allow string literal annotations like the string "Annotated[ClassVar[int]]". But since we'll no doubt have to support string-ized annotations even if PEP-649 is accepted, that's a pipe dream.

I think your suggestion for #1 seems reasonable.

For #2, in the case where typing has been imported and the annotation isn't a string, I assume it's simple enough to look inside the Annotated object and extract the InitVar (or ClassVar) object. I haven't delved in to Annotated object yet.

For #2, with a string annotation, it does look like it will get ugly. I'll have to spend some time looking at _MODULE_IDENTIFIER_RE. I'm guessing we'll need another re that looks for the same basic thing with "Annotated[" or "module.Annotated[" prepended and then look inside that. Or maybe one re to do both. I don't think we should support cases like:

from __future__ import annotations
myAnnotated = typing.Annotated

@dataclass
class Foo:
    a: myAnnotated[ClassVar[int]]

(That is, we won't recognize the string "myAnnotated[ClassVar[int]]" as an Annotated ClassVar.

Maybe we should also restrict it to "Annotated" or "typing.Annotated", but that would prevent someone from using "import typing as _typing", for example. This is why the current code accepts any module name, not strictly "typing". At least that's my recollection, more study is needed.

Or maybe it's time to give up and use typing.get_type_hints() or inspect.get_annotations(), but I suspect there are enough corner cases that will fail that we'll never get that to work right. Nested classes, anyone?

The whole runtime inspection of string-ized annotations is a mess.

On the "performance is important" comment: I'm not sure this is really an issue any more. There was some PEP that was supposed to speed up importing typing, and I never looked at the performance once it was merged. But then again, I'm not sure we want to always have dataclasses import typing, either. If a program doesn't use dataclasses that using the typing module, there's no sense importing it and enlarging the working set of modules.

I welcome any insights on any of these issues. I'm not a typing expert.

[GBeauregard]

Thanks for getting back so quickly.

Annotated is set up to be 'transparent' by default to typing.get_type_hints so in the case of using typing.py it can be made straightforward by chaining with typing.get_origin, I think.

I don't see any reasonable way to do the regex that allows Annotated to be renamed, so I agree your suggested restriction. I think the regex would be something like this:

^\s*(?:(?:\w+\s*\.)?\s*Annotated\[)?(?:\s*(\w+)\s*\.)?\s*(\w+)

I'm a bit worried people who are into Annotated annotations might be concerned about line length and more likely to rename Annotated, but I don't know if this is a realistic concern. And a part of me wants to say always importing typing.py should be okay since dataclasses was designed to work with type hints after all.

On the other hand, I'm also a bit worried that if we made dataclasses always import typing.py it would rub people the wrong way even if we profiled it and decided it was okay.

I'm going to spend some more time digesting the code tomorrow and try to decide if there's any major speed bumps to a full typing approach. I'll also look at import time and such.

[GBeauregard]

Or rather,

^\s*(?:(?:\w+\s*\.)?\s*Annotated\s*\[)?(?:\s*(\w+)\s*\.)?\s*(\w+)