Equality of typing.Literal depends on the order of arguments

[Dominik1123]

BPO	42345
Nosy	@gvanrossum, @ilevkivskyi, @miss-islington, @uriyyo, @Dominik1123, @Fidget-Spinner
PRs	bpo-42345: Fix three issues with typing.Literal parameters #23294 [3.9] bpo-42345: Fix three issues with typing.Literal parameters (GH-23294) #23335 bpo-42345: Fix hash implementation of typing.Literal #23383 bpo-42345: Add whatsnew for typing.Literal in 3.10 #23385 bpo-42345: Add whatsnew and versionchanged for typing.Literal in 3.9 #23386 [3.9] bpo-42345: Fix hash implementation of typing.Literal (GH-23383) #23408 [3.9] bpo-42345: Add whatsnew and versionchanged for typing.Literal in 3.9 (GH-23386) #23411

^{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 2020-11-20.17:11:24.442>
created_at = <Date 2020-11-13.14:11:55.639>
labels = ['type-bug', '3.8', '3.9', '3.10']
title = 'Equality of typing.Literal depends on the order of arguments'
updated_at = <Date 2021-10-13.10:26:49.678>
user = 'https://github.com/Dominik1123'

bugs.python.org fields:

activity = <Date 2021-10-13.10:26:49.678>
actor = 'kj'
assignee = 'none'
closed = True
closed_date = <Date 2020-11-20.17:11:24.442>
closer = 'gvanrossum'
components = []
creation = <Date 2020-11-13.14:11:55.639>
creator = 'Dominik V.'
dependencies = []
files = []
hgrepos = []
issue_num = 42345
keywords = ['patch']
message_count = 17.0
messages = ['380888', '380904', '381018', '381050', '381204', '381247', '381249', '381380', '381388', '381439', '381441', '381445', '381446', '381448', '381469', '403752', '403818']
nosy_count = 7.0
nosy_names = ['gvanrossum', 'alick', 'levkivskyi', 'miss-islington', 'uriyyo', 'Dominik V.', 'kj']
pr_nums = ['23294', '23335', '23383', '23385', '23386', '23408', '23411']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue42345'
versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']

[Dominik1123]

PEP-586 specifies that

Literal[v1, v2, v3]

is equivalent to

Union[Literal[v1], Literal[v2], Literal[v3]]

Since the equality of Unions doesn't take into account the order of arguments, Literals parametrized with multiple arguments should not be order dependent either. However they seem to:

    >>> Literal[1, 2] == Literal[2, 1]
    False

Compare with the equivalent form:

    >>> Union[Literal[1], Literal[2]] == Union[Literal[2], Literal[1]]
    True

In addition to that, the PEP specifies that nested Literals should be equivalent to the flattened version (https://www.python.org/dev/peps/pep-0586/#legal-parameters-for-literal-at-type-check-time). This section is titled "Legal parameters for Literal at type check time" but since the PEP doesn't specify runtime behavior differently, I think it makes sense to assume it is the same. It seems to be different though:

    >>> Literal[Literal[1, 2], 3]
    typing.Literal[typing.Literal[1, 2], 3]
    >>> Literal[Literal[1, 2], 3] == Literal[1, 2, 3]
    False

Also the flattening follows from the above definition Literal[v1, v2, v3] == Union[Literal[v1], Literal[v2], Literal[v3]] and the fact that Unions are flattened.

[gvanrossum]

Probably the implementation focused on static typing, not runtime checking.

Can you come up with a PR for a fix?

[Fidget-Spinner]

@guido and @ivan,

Should Literals de-duplicate values in its __args__ similar to Union? PEP-586 states that:

Literal[v1, v2, v3] is equivalent to Union[Literal[v1], Literal[v2], Literal[v3]]

Naturally, a Union de-duplicates values, so I was wondering if Literal should follow.

[gvanrossum]

Yeah, I think it makes sense to de-dupe args for Literal.

[miss-islington]

New changeset f03d318 by Yurii Karabas in branch 'master':
bpo-42345: Fix three issues with typing.Literal parameters (GH-23294)
f03d318

[miss-islington]

New changeset ac472b3 by Yurii Karabas in branch '3.9':
[3.9] bpo-42345: Fix three issues with typing.Literal parameters (GH-23294) (GH-23335)
ac472b3

[gvanrossum]

I recommend adding a whatsnew entry too. You can just add it to this issue. Interestingly you’ll probably need two separate ones, for 3.9 and 3.10. That would become two separate PRs for master, the 3.9 one to be backported.