gh-117557: Improve error message for 'C' format if given string is not 1 char long

[erlend-aasland]

• Issue: unicodedata: unicodedata.name() undocumented TypeError on empty string #117557

[ronaldoussoren]

LGTM

[serhiy-storchaka]

Is it really better than the current error message?

In any case, you cannot just change the error message in Argument Clinic. It should match the error message in PyArg_Parse.

There are other variants of the error message for the same error in ord, str.center, str and bytes formatting %c. If we change the error messages, it is better to consider all options and make them as consistent as possible.

[erlend-aasland]

Is it really better than the current error message?

Yes; it is clearer and more accurate.

There are other variants of the error message for the same error in ord, str.center, str and bytes formatting %c. If we change the error messages, it is better to consider all options and make them as consistent as possible.

They represent different APIs and require different error messages. Aligning them with this error message does not make sense to me.

[erlend-aasland]

In any case, you cannot just change the error message in Argument Clinic. It should match the error message in PyArg_Parse.

Yes, I think there is value in synchronising these.

[erlend-aasland]

I tried to address your concerns in fe02edd, Serhiy.

[serhiy-storchaka]

Thank you, it is better now. But note that other error message (few lines above) requires "a unicode character", not "a string containing exactly one unicode character". The latter looks as a more clumsy formulation of the former.

Ideally, the error message should contain the length of the string if it is not 1. But the PyArg_Parse code has limited facilities, so it needs more work.

Please look at more general image. Look how the affected functions are documented. Should the documentation be updated?

Look also at the 'c' format unit (bytes objects with length 1). Does it need changes?

[erlend-aasland]

Closed in favour of #117631.