Char type as a special length one string type

[yegle]

Cross post from typing repo: python/typing#510

The original issue contains some additional context, but is it feasible to implement a Char type as a special case of Text type?

[elazarg]

I think this is similar (in challenges and special-casing) to having a singleton type. (Imagine unary alphabet; for most purposes mypy doesn't know it's not). But it's less general and requires understanding of len for flow-dependent checking. If we can do this, we might as well add singleton types, and a special "AnyChar" value. Then Char is Singleton[AnyChar].

[JukkaL]

I've sometimes wondered about having a character type, but haven't yet found very convincing use cases.

One option would be to implement this as a separate string-like type with a promotion from Char to Text (similar to the promotion from int to float that we already have). Both would be represented at runtime by the same type. A single-character string literal would have type Char (maybe only in a Char type context, though).

However, there are probably simpler ways to catch accidentally using str when Iterable[str] or Sequence[str] is expected. For example, we could have a strictness option that would flag these. If we had such option, I'd probably turn it on in codebases that I work on, for sure.

[JukkaL]

Literal['x'] can sometimes be used for this use case. More generally, there is no concrete proposal that seems practical, so I'm closing the issue.

[Hibou57]

There is also a common misconception about what a character is. There are code points and graphemes. What one sees in a text editor, are graphemes, what a computer sees in a file, are code points. Some graphemes are made of two code points, a code point for a base glyph and a code point for an overlayed glyph.

As far I experienced it, Python deals with code point, so a grapheme made of two glyphs, will be a string of length 2.

The best to deal with code points individually, is to deal with their integer value which can be retrieve with the ord() function.

I would rather say the issue is that Python does not provide a built-in iterator to iterate over a string as a sequence of code point integer values, but one can define such an iterator easily. I also believe there is no Python function to split a string into graphemes, by the way (unless i’m wrong).