v6.0.0

Migration: FloatPointCompatible now uses an associated scalar type

FloatPointCompatible no longer takes the scalar as a generic parameter. Implementations should define type Scalar instead.

Before:

impl FloatPointCompatible<f32> for MyPoint {
    fn from_xy(x: f32, y: f32) -> Self { ... }
    fn x(&self) -> f32 { ... }
    fn y(&self) -> f32 { ... }
}

After:

impl FloatPointCompatible for MyPoint {
    type Scalar = f32;

    fn from_xy(x: f32, y: f32) -> Self { ... }
    fn x(&self) -> f32 { ... }
    fn y(&self) -> f32 { ... }
}

Public float overlay APIs were simplified accordingly. Types that previously carried both point and scalar generics now generally use only the point type, with scalar values expressed as P::Scalar.

v5.0.0

Starting with 5.0.0, i_overlay follows SemVer release cycles.

Changes

• Switched release policy to SemVer.
• Added cargo-semver-checks to CI.

v4.5.0

Changes

• Changed: Reworked buffering algorithm. Now it's relay on default boolean overlay graph. Old legacy buffering overlay graph removed. New algorithm must be more precision stable and safe.
• Added: Unit tests.

v4.2.1

Changes

• Changed: License updated from MIT to dual MIT OR Apache-2.0.
• Added: Integration tests.

v4.2.0

Changes

• Added OGC-valid output mode to split composite holes into valid contours.
• Added fixed-scale/grid_size float overlay support to make operations more stable and predictable.

v4.0.0

Added

• no_std support
  std math replaced with libm.

Modified

• Memory-efficient API
  FloatOverlay and Overlay use a reusable buffer to reduce allocation overhead, improving performance by 5–15% in repeated operations.
  OverlayGraph is now a reference view (no longer owns geometry data)

• Optional multithreading
  allow_multithreading is now an optional feature and disabled by default.

v3.3.0

Fixed

API Design: options were mistakenly present in both the constructor and the custom function. Now they are only in the constructor.

v3.2.0

Modified

Area type changed: usize -> u64
Using usize caused incorrect results on WebAssembly.

v3.1.1

Added

IntOverlayOptions: options for integer boolean operations.

• preserve_input_collinear: Preserve collinear points before operations.
• output_direction: Desired output winding order.
• preserve_output_collinear: Preserve collinear points after operations.
• min_output_area: Minimum area threshold for output contours.

OverlayOptions<T>: options for float boolean operations.

• preserve_input_collinear: Preserve collinear points before operations.
• output_direction: Desired output winding order.
• preserve_output_collinear: Preserve collinear points after operations.
• min_output_area: Minimum area threshold for output contours.
• pub clean_result: If true, the result will be cleaned from precision-related issues (moved from ContourFilter)

❌ Removed

ContourFilter<T>: Previously used for controlling output simplification and area threshold; now replaced by the more unified OverlayOptions.

3.0.1

Changed

Starting from this version:

• Outer contours are returned in counter-clockwise order
• Hole contours are in clockwise order

This aligns with the standard mathematical convention and improves compatibility with other geometry libraries.
Now it also can be adjustable with enum ContourDirection

API Cleanup

Advanced methods have been renamed for clarity and now use the *_custom suffix:

• overlay_custom
• extract_shapes_custom
• slice_custom_by
  ...and others

The slicing algorithm has been reworked and became more stable.