Lib.rs

This is a preliminary "does this look okay to use" code review, writing down everything I noticed -- I have not yet attempted to use the library.

Build

Good.

• Has no_std support, in the conventional fashion of a std feature that can be disabled.
• Depends only on common interop libraries, and optionally libm. All dependencies are optional except num-traits.

API design

Dubious in places.

• Rect::inner_rect() contains a debug_assert! that the size of the result is not negative, yet this is not an invariant of the type, and the potential panic is not documented.
• The approxord module contains no items that match its module docs, just max() and min() for PartialOrd rather than only Ord.

The following are “this doesn't exist” rather than “this exists but is bad”:

• Angle and Rotation2D type are specifically radians. This prevents some opportunities for exact arithmetic on fractions or multiples of whole turns (because 360 is a rational number and 2π is not).
• There is no Vector4D type, only the specific-purpose HomogeneousVector.

Documentation

• Many functions without docs or with not very specific docs.
• cast() methods have an un-track_callered panic if the underlying num_traits::NumCast fails. On the ones which have documentation at all (e.g. Length), the description of cast() and try_cast() as "Cast" and "Fallible cast" respectively might lead the reader to think that the former is an infallible conversion, not a panicking one.

Code

• Clippy lint directives in an archaic style from before they were namespaced: #![cfg_attr(feature = "cargo-clippy", allow(just_underscores_and_digits))]
• Superfluous use mint; items.
• approxeq.rs: Default epsilons are a dubious concept in the first place, but particularly, using 1e-6 for f64 as well as f32 seems a bit lax.
• rigid.rs: "All matrix multiplication in this module is in row-vector notation," which seems gratuitously inconsistent with most common notation and with the transform2d and transform3d modules. But perhaps I have misunderstood.

Pleasantly unsurprising. It's just math.

Only in debcargo (unstable). Changelog:

• Team upload.
• Package euclid 0.22.7 from crates.io using debcargo 2.6.0

Packaged for Guix (crates-graphics)