char::{to_uppercase, to_lowercase} are broken and should not be marked stable

[niconii]

The description of char::to_uppercase reads:

The case-folding performed is the common or simple mapping: it maps one Unicode codepoint to its uppercase equivalent according to the Unicode database. The additional [SpecialCasing.txt] is not yet considered here, but the iterator returned will soon support this form of case folding.

Similarly, char::to_lowercase reads:

The case-folding performed is the common or simple mapping. See to_uppercase() for references and more information.

Right now, char::to_uppercase and char::to_lowercase disregard graphemes whose lowercase codepoints do not map one-to-one with their uppercase codepoints. For example, U+FB02 fl LATIN SMALL LIGATURE FL, a single codepoint, maps to two characters: 'F' and 'L'. Yet, when we try converting it to uppercase in Rust...

fn main() {
    let uppercase_flavor = "flavor".chars()
        .flat_map(char::to_uppercase)
        .collect::<String>();

    println!("{}", uppercase_flavor);
}

...rather than the expected "FLAVOR", we get "flAVOR", an incorrect result.

As the description of char::to_uppercase explains, this behavior is expected to change. However, currently char::to_uppercase, char::to_lowercase, and their associated iterators std::char::ToUppercase and std::char::ToLowercase are all marked stable, yet fixing this behavior would be a breaking change.

[alexcrichton]

Thanks for the report! This sort of change is definitely planned, however, and has yet to be implemented. The switch from returning just a plain char to returning an iterator of characters reflects this. A minor update such as this will not be considered a breaking change, especially because of the way the documentation is currently worded. As a result I'm going to close this issue for now in favor of #25800, which reflects the current plan.