Tracking Issue for `Duration::as_millis_{f64,f32}`

Feature gate: #![feature(duration_millis_float)]

This is a tracking issue for

• Duration::as_millis_f64
• Duration::as_millis_f32

Public API

impl Duration {
    /// Returns the number of milliseconds contained by this `Duration` as `f64`.
    ///
    /// The returned value does include the fractional (nanosecond) part of the duration.
    ///
    /// # Examples
    /// ```
    /// use std::time::Duration;
    ///
    /// let dur = Duration::new(2, 67_890_000);
    /// assert_eq!(dur.as_millis_f64(), 2678.9);
    /// ```
    pub const fn as_millis_f64(self);

    /// Returns the number of milliseconds contained by this `Duration` as `f32`.
    ///
    /// The returned value does include the fractional (nanosecond) part of the duration.
    ///
    /// # Examples
    /// ```
    /// use std::time::Duration;
    ///
    /// let dur = Duration::new(2, 67_890_000);
    /// assert_eq!(dur.as_millis_f32(), 2678.9);
    /// ```
    pub const fn as_millis_f32(self);
}

Steps / History

• ACP: https://github.com/rust-lang/libs-team/issues/349
• Implementation: https://github.com/rust-lang/rust/pull/122479
• Related lint: https://github.com/rust-lang/rust-clippy/pull/12539
• Final comment period (FCP)^1
• Stabilization PR

Unresolved Questions

• The implementation is using floating-point division; would we be ok with changing it to multiplication-by-reciprocal at the cost of a ½ULP but being faster? as_secs_f32 uses division, so maybe not.

Review the public API and the linked implementation PR first, then read the feature-lifecycle stabilization guide referenced by the issue. The remaining work is the final comment period and a stabilization PR; the unresolved question about floating-point division versus multiplication by a reciprocal also needs resolution.