Expose `FloatErrorKind` from private `core::num::dec2flt` to public `core::num::FloatErrorKind`

[tisonkun]

Proposal

Problem statement

core::num::ParseIntError can expose its kind via:

impl ParseIntError {
    /// Outputs the detailed cause of parsing an integer failing.
    #[must_use]
    #[rustc_const_stable(feature = "const_int_from_str", since = "1.82.0")]
    #[stable(feature = "int_error_matching", since = "1.55.0")]
    pub const fn kind(&self) -> &IntErrorKind {
        &self.kind
    }
}

However, core::num::ParseFloatError cannot expose its kind. Neither a kind method nor the FloatErrorKind enum is publicly available.

This is unnecessary unalignment IMO, so I propose to expose FloatErrorKind and add ParseFloatError::kind method.

Motivating examples or use cases

Described as above. I have a personal project that handle these two errors:

impl From<ParseIntError> for ErrorKind {
    fn from(err: ParseIntError) -> Self {
        ErrorKind::Other(match err.kind() {
            IntErrorKind::InvalidDigit => ...,
            IntErrorKind::PosOverflow => ...,
            IntErrorKind::NegOverflow => ...,
            _ => ...,
        })
    }
}

impl From<ParseFloatError> for ErrorKind {
    fn from(err: ParseFloatError) -> Self {
        ErrorKind::Other(match err.kind() { /* compile error, no `kind` method and the kind enum is private */})
    }
}

Solution sketch

1. Mark core::num::dec2flt::FloatErrorKind as pub, properly expose it cascadingly to core::num.
2. Add ParseFloatError::kind to expose the error kind, just as ParseIntError::kind does.

Alternatives

N/A

Links and related work

• ParseIntError and ParseFloatError does not allow to introspect the cause rust#22639

What happens now?

This issue contains an API change proposal (or ACP) and is part of the libs-api team feature lifecycle. Once this issue is filed, the libs-api team will review open proposals as capability becomes available. Current response times do not have a clear estimate, but may be up to several months.

Possible responses

The libs team may respond in various different ways. First, the team will consider the problem (this doesn't require any concrete solution or alternatives to have been proposed):

• We think this problem seems worth solving, and the standard library might be the right place to solve it.
• We think that this probably doesn't belong in the standard library.

Second, if there's a concrete solution:

• We think this specific solution looks roughly right, approved, you or someone else should implement this. (Further review will still happen on the subsequent implementation PR.)
• We're not sure this is the right solution, and the alternatives or other materials don't give us enough information to be sure about that. Here are some questions we have that aren't answered, or rough ideas about alternatives we'd want to see discussed.

[programmerjake]

I don't think FloatErrorKind is in the shape to be fully exposed how it is since it currently only contains Empty and Invalid.

So, I think either we'd have to mark it #[non_exhaustive] and have just the Empty variant be stable, or we'd want to drastically improve the quality of errors returned, e.g. adding variants based on https://docs.rs/lexical-parse-float/1.0.6/lexical_parse_float/enum.Error.html

[tisonkun]

mark it #[non_exhaustive]

I think this is the way to go. Sorry not to include it in the description. IntErrorKind is marked as #[non_exhaustive] also.

https://docs.rs/lexical-parse-float/1.0.6/lexical_parse_float/enum.Error.html

This is also a good point that can be investigated during the implementation or as a follow-up.

[Alexhuszagh]

I agree with making it public, only have Empty be stable, and have it #[non_exhaustive]. As an enhancement, I think we could have a few simple cases that would likely cover most cases:

1. . is parsed (some new variant of empty)
2. 123.45b (trailing digits, invalid but checks if any digits have been parsed)
3. abc134 (invalid)

And we would have 4 variants to handle most use-cases, minimal code changes, as well as room for enhancements as needed. Let me know if I can help in any way.

[programmerjake]

it would be nice if we could have the error maybe contain a parsed float and how much was parsed if some prefix of the input is a syntactically valid float -- like strtod in C. AFAICT this is why all the lexical_parse_float::Error variants contain usize. Though, because we have ParseFloatError, those can be fields of ParseFloatError rather than having to stuff them in FloatErrorKind