Python modifies unicode identifiers?

Question

Python 3.8 supports using a limited set of non-ASCII Unicode characters in identifiers. So, it seems that it is valid to use as a character in an identifier.

However, something is wrong...

Problem

def f():
    print(f'{=}')

f(1)
f(=2)
f(**{'': 3})

The first two calls are fine, but the third fails:

=1
=2
Traceback (most recent call last):
  File "sigma.py", line 24, in <module>
    f(**{'': 3})
TypeError: f() got an unexpected keyword argument ''

Analysis

Let's see what is actually going on:

def f2(**kw):
    for name, value in kw.items():
        print(f'{name}={value}     {ord(name)=}')
f2(=2)
f2(**{'': 3})

It prints:

?=2     ord(name)=931
=3     ord(name)=120506

I called it with both times, but it was changed to the very similar simpler ? in the first call.

It seems that an argument named (U+1D6BA) is implicitly renamed to ? (U+03A3), and in every call to the function, argument is also implicitly renamed to ?, except if it is passed as **kwargs.

The Questions

Is this a bug? It does not look like it is accidental. Is it documented? Is there a set of true characters and a list of alias characters available somewhere?

Answer 1

我认为这是由于 Python 处理字符的方式造成的。
\n如果您使用提供的西格玛字母之一设置变量： \xce\xa3 或 ，您也可以使用另一个来访问它。知道这两个片段都有效：

\n

>>> \xce\xa3 = 5\n>>> \n5\n

\n

>>>  = 5\n>>> \xce\xa3\n5\n

\n

您可以看到globals()它被分配给 \xce\xa3 (ord: 931)
\n我的猜测是 Python 在执行变量查找之前修改了该字符。
\n类似的讨论，由我在 github/wtfpython 中发布

\n