Unicode support about combine HOT 3 CLOSED

For classes only, yes. For normalisation, no, one would have to write a new parser using the unicode-normalization crate, because a token here means a single codepoint (char), but normalisation needs to see multiple-codepoint sequences.

In fact I tried—and failed (#85)—at similar thing with graphemes and unicode-segmentation yesterday.

And then it depends on how generic you'd need the parser to be. unicode-normalization works on &str, so if you have that, you can use it directly, but if you have some other kind of stream, you have to collect and try normalising.

from combine.

No, you will have to look into some other crate for that. You can then use satisfy(|token| predicate(token)).expected("Expected value") to create a custom parser for whatever predicate you want (digit, letter etc are all implemented like this)

from combine.

To get regex parsers working I added FullRangeStream which lets one get retrieve a view into the entire parsers input.
Since unicode parsers require &str then this can be used to write parsers generic over the input stream (translating the example in #85 (comment).

pub fn grapheme<'a, I>(input: I) -> ParseResult<&'a str, I>
    where I: FullRangeStream<Range = &'a str>
{
    let mut iter = input.range().graphemes();
    match iter.next() {
        Some(g) => {
            let len = input.range().len() - iter.as_str().len();
            take(len).parse_stream(input)
        }
        None => Err(ParseError::end_of_input()),
    }
}

from combine.