Decide on names for icu_datetime and icu_calendar errors over FFI

[sffc]

icu_datetime and icu_calendar have many new, small error enums. We should discuss their naming conventions.

One example which isn't super great is that icu_calendar::ParseError is currently exported as CalendarParseError, even though it is the error resulting from parsing an IXDTF string to a Date, Time, or DateTime. (https://github.com/unicode-org/icu4x/pull/5260)

[sffc]

I want to save this discussion until after neo datetime is fully landed so that we can see exactly which error enums we're having to name in FFI.

[sffc]

Part of the 2.0 FFI audit

[sffc]

@robertbastian says he's fine with names so long as they are targeted/descriptive as opposed to reliant on the crate name.

[Manishearth]

This is blocked on https://github.com/unicode-org/icu4x/issues/5940

We don't seem to need a discussion yet, Shane plans to make a PR when ready and we can discuss if there is controversy.

[sffc]

Here's what we're looking at:

#[derive(Debug, PartialEq, Eq)]
#[repr(C)]
#[diplomat::rust_link(icu::calendar::RangeError, Struct, compact)]
#[diplomat::rust_link(icu::calendar::DateError, Enum, compact)]
#[cfg(any(feature = "datetime", feature = "timezone", feature = "calendar"))]
pub enum CalendarError {
    Unknown = 0x00,
    OutOfRange = 0x01,
    UnknownEra = 0x02,
    UnknownMonthCode = 0x03,
}

#[derive(Debug, PartialEq, Eq)]
#[repr(C)]
#[diplomat::rust_link(icu::calendar::ParseError, Enum, compact)]
#[diplomat::rust_link(icu::time::ParseError, Enum, compact)]
#[cfg(any(feature = "datetime", feature = "timezone", feature = "calendar"))]
pub enum CalendarParseError {
    Unknown = 0x00,
    InvalidSyntax = 0x01,
    OutOfRange = 0x02,
    MissingFields = 0x03,
    UnknownCalendar = 0x04,
}

#[derive(Debug, PartialEq, Eq)]
#[diplomat::rust_link(icu::time::zone::InvalidOffsetError, Struct, compact)]
#[cfg(any(feature = "datetime", feature = "timezone"))]
pub struct TimeZoneInvalidOffsetError;

#[derive(Debug, PartialEq, Eq)]
#[repr(C)]
#[diplomat::rust_link(icu::datetime::DateTimeFormatterLoadError, Enum, compact)]
#[diplomat::rust_link(icu::datetime::pattern::PatternLoadError, Enum, compact)]
#[diplomat::rust_link(icu_provider::DataError, Struct, compact)]
#[diplomat::rust_link(icu_provider::DataErrorKind, Enum, compact)]
pub enum DateTimeFormatterLoadError {
    Unknown = 0x00,

    InvalidDateFields = 0x8_01,
    UnsupportedLength = 0x8_03,
    ConflictingField = 0x8_09,
    FormatterTooSpecific = 0x8_0A,

    DataMarkerNotFound = 0x01,
    DataIdentifierNotFound = 0x02,
    DataInvalidRequest = 0x03,
    DataInconsistentData = 0x04,
    DataDowncast = 0x05,
    DataDeserialize = 0x06,
    DataCustom = 0x07,
    DataIo = 0x08,
}

#[cfg(feature = "datetime")]
#[diplomat::rust_link(icu::datetime::MismatchedCalendarError, Struct)]
pub struct DateTimeMismatchedCalendarError {
    pub this_kind: CalendarKind,
    pub date_kind: DiplomatOption<CalendarKind>,
}

#[cfg(feature = "datetime")]
#[derive(Debug, PartialEq, Eq)]
#[repr(C)]
#[diplomat::rust_link(
    icu::datetime::unchecked::FormattedDateTimeUncheckedError,
    Enum,
    compact
)]
pub enum DateTimeWriteError {
    Unknown = 0x00,
    MissingTimeZoneId = 0x01,
    MissingTimeZoneNameTimestamp = 0x02,
    MissingTimeZoneVariant = 0x03,
}

[sffc]

DateTimeFormatterLoadError probably has more variants than are reachable. Do we care?

I don't really like the name of CalendarParseError. It applies to the parsing in both the icu_calendar and icu_time crates.

The others look plausible.

@robertbastian

[sffc]

Brief discussion with @robertbastian:

• Debug-assert the errors that we think are not reachable
• Rename CalendarParseError to Rfc9557ParseError