Convex MCP server

//! Helpers for working with floating point numbers. use std::cmp::Ordering; /// Is a floating point number an integer? Note that we do not consider negative /// zero an integer, since we want there to be an injection between the subset /// of `is_integral` floating point numbers and the integers. #[allow(clippy::float_cmp)] pub fn is_integral(n: f64) -> Option<i64> { // `n == n.trunc()` isn't optimized very well by the compiler. If this is // ever important, we just need to check that 1) the f64 is normal and 2) // the exponent is greater than (64 - the number of trailing zeros of the // mantissa) to ensure we shift the decimal point past all of the mantissa's // digits. Note also that subnormal numbers are never integers. // See https://stackoverflow.com/questions/26341494 for more color. let truncated = n.trunc(); if !is_negative_zero(n) && !n.is_infinite() && truncated == n { Some(truncated as i64) } else { None } } /// Is a floating point number native zero? pub fn is_negative_zero(n: f64) -> bool { matches!(n.total_cmp(&-0.0), Ordering::Equal) }