Repository navigation
Add types.Self #90827
Description
| BPO | 46669 |
|---|---|
| Nosy | @gvanrossum, @rhettinger, @JelleZijlstra |
| Superseder |
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 = <Date 2022-02-07.18:09:47.676>
created_at = <Date 2022-02-06.23:12:51.840>
labels = ['3.11']
title = 'Add types.Self'
updated_at = <Date 2022-02-08.05:25:38.097>
user = 'https://github.com/rhettinger'bugs.python.org fields:
activity = <Date 2022-02-08.05:25:38.097>
actor = 'rhettinger'
assignee = 'none'
closed = True
closed_date = <Date 2022-02-07.18:09:47.676>
closer = 'rhettinger'
components = []
creation = <Date 2022-02-06.23:12:51.840>
creator = 'rhettinger'
dependencies = []
files = []
hgrepos = []
issue_num = 46669
keywords = []
message_count = 5.0
messages = ['412682', '412684', '412772', '412786', '412813']
nosy_count = 3.0
nosy_names = ['gvanrossum', 'rhettinger', 'JelleZijlstra']
pr_nums = []
priority = 'normal'
resolution = 'duplicate'
stage = 'resolved'
status = 'closed'
superseder = '46534'
type = None
url = 'https://bugs.python.org/issue46669'
versions = ['Python 3.11']
Activity
rhettinger commentedon Feb 6, 2022
Typeshed now has a nice self-describing type variable to annotate context managers:
It would be nice to have that in the standard library types module as well.
JelleZijlstra commentedon Feb 6, 2022
PEP-673 (which was accepted) adds typing.Self already. bpo-46534 tracks implementing it.
rhettinger commentedon Feb 7, 2022
On a related note, is this the correct way to annotate __exit__?
JelleZijlstra commentedon Feb 7, 2022
The conventional way is to write
But there are two invariants about how __exit__ works in practice that this doesn't capture:
I believe these invariants are always true in 3.11 thanks to Irit's work, but previously there could be rare circumstances where they didn't hold. You probably know the details better than I do.
We could support the first invariant by using two overloads, one where the arguments are all None and one where they aren't. I don't think that's generally worth doing though: it's a lot of verbose code and doesn't fix many real problems.
Your suggested signature looks like it's trying to support the second invariant, but it doesn't quite: if the types don't match, the type checker will just set T to the common base type of the two arguments.
rhettinger commentedon Feb 8, 2022
Is there a way to write the second invariant correctly?