Extend FromStr by from_{str,bytes}{,_initial}

[mahkoh]

• Right now it is impossible to parse strings of the form [0-9]+.+ without practically rewriting the parser or using the correct libc function (strtol).
• Most of the time the UTF-8 property is irrelevant. Most things parse bytes.

trait FromBytes<T> {
    /// Tries to convert the initial portion of `b` to `T` and returns the number of bytes
    /// consumed.
    fn from_bytes_initial(b: &[u8]) -> Option<(T,uint)>;

    /// Tries to convert `b` to `T`.
    fn from_bytes(b: &[u8]) -> Option<T> {
        match FromBytes::from_bytes_initial(b) {
            Some((t, n)) if n == b.len() => Some(t),
            _ => None,
        }
    }

    /// Tries to convert the initial portion of `s` to `T` and returns the number of bytes
    /// consumed.
    fn from_str_initial(s: &str) -> Option<(T,uint)> {
        FromBytes::from_bytes_initial(s.as_bytes())
    }

    /// Tries to convert `s` to `T`.
    fn from_str(s: &str) -> Option<T> {
        FromBytes::from_bytes(s.as_bytes())
    }
}

[aturon]

cc me

[huonw]

cc #6220 (at least, this sort of API is mentioned there and would be useful for it).

I don't know if want from_bytes to be the default so deeply, since things that do want to parse unicode are then required to remember to override everything, at the expense of a pile of from_utf8 calls (especially bad for something recursive). Maybe this doesn't matter much in practice.

[mahkoh]

There are several places in the stdlib that already roll their own parsers because of this.