F821 triggers when using non-ASCII characters

[trag1c]

from time import time_ns

def time_µs() -> int:
	return time_ns() // 1000

def main() -> None:
	print(time_µs())

if __name__ == "__main__":
	main()

test.py:7:9: F821 Undefined name `time_μs`

(it's obviously fine with a name like time_us)

[charliermarsh]

Fascinating.

[charliermarsh]

Oddly I'm not seeing this error when running on my machine 🤔

[trag1c]

Oddly I'm not seeing this error when running on my machine 🤔

Do you need any more details?

[charliermarsh]

Could it make a difference if you uploaded the exact file, rather than pasting the snippet?

[trag1c]

Could it make a difference if you uploaded the exact file, rather than pasting the snippet?

Unfortunately, I don't have the exact file anymore. I didn't actually run this snippet until now, and oddly it also seems to work for me in this case.

It complains about it in this example though:

def time_µs():
    ...

def main():
    time_μs()

[charliermarsh]

Oh cool, that repros for me too. Thanks!

[charliermarsh]

Really don't understand this one.

It seems to work fine if you indent with tabs, but not with spaces?

def time_µs() -> int:
	...

def main() -> None:
	time_µs()

[charliermarsh]

Wait, no, that's wrong -- the tabs are irrelevant.

[charliermarsh]

I think there's something strange in how the time_µs is written in the first example. Notice that my editor doesn't like it either:

[charliermarsh]

Yeah, I think these are actually using different characters? In foo.py above, the first representation of time_us uses:

116
105
109
101
95
194
181
115

The second uses:

116
105
109
101
95
206
188
115

The difference is [206, 188] vs. [194, 181], which I guess represent slightly different characters?

string_utf8_lossy: μ
string_utf8_lossy: µ

[charliermarsh]

So, I think this is correct?

[charliermarsh]

Sigh, yes, the former ([206, 188]) is Greek Small Letter Mu, while the latter ([194, 181]) is Micro Sign. You can see that the UTF-8 Encoding hex digits match up.

[trag1c]

Sorry for wasting your time, looks like it's VSCode that treats these two as the same character, e.g. when highlighting or using autocompletion (that's probably when it inserted a different character? 🤨).

Anyways great project! ❤️

[charliermarsh]

Haha no, all good, I was totally stumped! I learned a bit about debugging this stuff anyway.

Thank you for the kind words :)

[ekzhang]

(Off-topic) By the way, Rust has a built-in lint for this, just tried compiling the code in this playground.

[covracer]

Hopefully #4430 will help prevent micro/mu mixups

I was also surprised by PEP 3131 edge cases.