namedtuple should use NFKD to find duplicate members

[JohnCooke]

BPO	33880
Nosy	@rhettinger, @ericvsmith, @isidentical

^{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 2019-12-08.17:51:19.375>
created_at = <Date 2018-06-16.17:12:33.255>
labels = ['3.8', 'type-bug', 'library', 'invalid']
title = 'namedtuple should use NFKD to find duplicate members'
updated_at = <Date 2019-12-08.17:51:19.144>
user = 'https://bugs.python.org/JohnCooke'

bugs.python.org fields:

activity = <Date 2019-12-08.17:51:19.144>
actor = 'rhettinger'
assignee = 'none'
closed = True
closed_date = <Date 2019-12-08.17:51:19.375>
closer = 'rhettinger'
components = ['Library (Lib)']
creation = <Date 2018-06-16.17:12:33.255>
creator = 'John Cooke'
dependencies = []
files = []
hgrepos = []
issue_num = 33880
keywords = []
message_count = 7.0
messages = ['319763', '319765', '319767', '319771', '319851', '357996', '358013']
nosy_count = 4.0
nosy_names = ['rhettinger', 'eric.smith', 'John Cooke', 'BTaskaya']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue33880'
versions = ['Python 3.8']

[JohnCooke]

from collections import namedtuple
# create a namedtuple whose members are:
# 00B5;MICRO SIGN;Ll;
# 03BC;GREEK SMALL LETTER MU;Ll
# these are both legal identifier names
names = ['\u00b5', '\u03bc']
for name in names:
    assert name.isidentifier()
mu = namedtuple('mu', names)

# should have raised ValueError, but instead get SyntaxError

[ericvsmith]

Not that it really matters to this issue, but here's how dataclasses and attrs deal with this:

dataclasses has the same issue, via make_dataclass().

attrs gives a syntax error with your field names, but interestingly this succeeds:

>> Foo = attr,make_class('Foo', ['a', 'b', 'a'])
>> Foo(1, 3)

I'll open a corresponding issue for dataclasses.
Foo(a=1, b=3)

[ericvsmith]

See bpo-33881 for the corresponding dataclasses issue.

[ericvsmith]

Actually, should this be NKFC?

From https://docs.python.org/3.6/reference/lexical_analysis.html#identifiers :

"All identifiers are converted into the normal form NFKC while parsing; comparison of identifiers is based on NFKC."

[rhettinger]

Of late, people have been very concerned with namedtuple construction time, so I'm disinclined to add this overhead for something that doesn't seem to have been a problem in the real world.

Also, the SyntaxError seems reasonable. That is the same error given by a regular function definition or types.SimpleNamespace:

    >>> def f(µ=1, μ=2):
    	    pass
    SyntaxError: duplicate argument 'μ' in function definition

    >>> SimpleNamespace(µ=1, μ=2)
    SyntaxError: keyword argument repeated

[isidentical]

Hey @rhettinger, what is the status of this issue? Is there a consensus about fixing it or can this issue be closed?

[rhettinger]

Marking as closed for the reasons listed above.

Other thoughts:

• In the class form with types.NamedTuple, no error is raised.
• In the function form with collections.namedtuple, the various
  TypeErrors and ValueErrors are just courtesy checks
  intended to provide slightly nicer error messages when possible.
  It wasn't the goal to upstage all possible syntax errors.