raygeo 0.8.0

A 2D/3D geometry library for CAD/CAM applications.

raygeo

A high-performance 2D/3D geometry library for Python, built in Rust with PyO3.

raygeo provides vector path construction, polygon boolean operations, curve fitting, path transformations, and geometric queries — all backed by a native Rust extension.

Concave hull, arc fitting, nesting, and raster power modulation

Installation

pip install raygeo

Requires Python 3.10+ and a compatible platform (Linux, Windows, macOS Intel, or macOS Apple Silicon). Pre-compiled wheels are available on PyPI.

Quick Start

Building Paths

The Geometry class is the core abstraction. It stores a vector path as a sequence of move, line, arc, and cubic Bezier commands. All mutating methods return self for chaining:

from raygeo.geo import Geometry

# Create a 10x10 square
g = Geometry()
g.move_to(0, 0)
g.line_to(10, 0)
g.line_to(10, 10)
g.line_to(0, 10)
g.close_path()

print(g.area())    # 100.0
print(g.rect())    # (0.0, 0.0, 10.0, 10.0)
print(g.is_closed())  # True

Builder methods can be chained:

g = Geometry()
g.move_to(0, 0).line_to(10, 0).line_to(10, 10).line_to(0, 10).close_path()

You can also create paths from point lists:

triangle = Geometry.from_points([(0, 0), (10, 0), (5, 8.66)])

Arcs and Bezier Curves

g = Geometry()
g.move_to(0, 0)
g.arc_to(10, 0, i=5, j=0, clockwise=False)  # semicircular arc
g.close_path()

# Bezier curves
g2 = Geometry()
g2.move_to(0, 0)
g2.bezier_to(10, 0, c1x=3, c1y=5, c2x=7, c2y=5)

# Convert arcs to Bezier curves (for non-uniform scaling)
g3 = Geometry()
g3.move_to(0, 0)
g3.arc_to_as_bezier(10, 0, i=5, j=0)
g3.upgrade_to_scalable()

Path Analysis

print(g.distance())       # total path length
print(g.area())           # signed enclosed area
print(g.rect())           # bounding box (x_min, y_min, x_max, y_max)
print(g.is_closed())      # path closure check
print(g.segments())       # split into sub-paths

# Find closest point on path
result = g.find_closest_point(5, 5)  # (segment_index, t, (x, y)) or None

# Point and tangent at parameter t on a segment
point = g.get_point_at(segment_index=0, t=0.5)
tangent = g.get_tangent_at(segment_index=0, t=0.5)

Transformations

All transformation methods mutate the geometry in place and return self, allowing chaining. Use .copy() first if you need to preserve the original:

import numpy as np
from raygeo.geo import Geometry

g = Geometry.from_points([(0, 0), (10, 0), (10, 10), (0, 10)])

# Offset (grow/shrink) — mutates in place
g.grow(2.0)   # offset outward by 2 units
print(g.area())  # 144.0

# Use .copy() to preserve the original
original = Geometry.from_points([(0, 0), (10, 0), (10, 10), (0, 10)])
shrunk = original.copy()
shrunk.grow(-1.0)  # offset inward by 1 unit

# Affine transform (4x4 matrix) — mutates in place
matrix = [
    [1, 0, 0, 5],  # translate x by 5
    [0, 1, 0, 3],  # translate y by 3
    [0, 0, 1, 0],
    [0, 0, 0, 1],
]
g.transform(matrix)

# Map geometry into a frame — mutates in place
g.map_to_frame(
    origin=(0, 0),
    p_width=(100, 0),
    p_height=(0, 100),
)

g.flip_x()  # negate all x coordinates
g.flip_y()  # negate all y coordinates

# Chaining is possible since all methods return self
g2 = Geometry.from_points([(0, 0), (10, 0), (10, 10), (0, 10)])
g2.transform(matrix).flip_x().grow(1.0)

Contour Operations

All contour methods mutate the geometry in place and return self:

# Split into separate closed contours (returns list, does not mutate)
contours = g.split_into_contours()

# Split into disconnected components (returns list, does not mutate)
components = g.split_into_components()

# Separate holes from solids (returns tuple, does not mutate)
inner, outer = g.split_inner_and_outer_contours()

# Normalize winding orders — mutates in place
g.normalize_winding_orders()

# Filter to only external contours — mutates in place
g.filter_to_external_contours()

# Remove shared edges between sub-paths — mutates in place
g.remove_inner_edges()

Polygon Operations

The geo.shape.polygon submodule provides polygon-specific operations powered by Clipper2:

from raygeo.geo import Geometry
from raygeo.geo.shape.polygon import (
    get_polygon_area,
    get_polygon_bounds,
    offset_polygon,
    get_polygons_union,
    get_polygons_intersection,
    get_polygons_difference,
    is_point_inside_polygon,
    polygons_intersect,
    get_polygon_convex_hull,
)

square = [(0, 0), (10, 0), (10, 10), (0, 10)]
circle_approx = [(5 + 5 * math.cos(a), 5 + 5 * math.sin(a))
                 for a in [i * math.pi / 20 for i in range(40)]]

get_polygon_area(square)                # 100.0
get_polygon_bounds(square)              # (0.0, 0.0, 10.0, 10.0)
is_point_inside_polygon((5, 5), square) # True

# Boolean operations
union = get_polygons_union([square, circle_approx])
intersection = get_polygons_intersection(square, circle_approx)
difference = get_polygons_difference(square, circle_approx)

# Offset
inflated = offset_polygon(square, 2.0)

# NumPy variants are also available (suffixed with _numpy)
import numpy as np
sq_np = np.array(square)
get_polygon_area(sq_np)  # also works with numpy arrays

Curve Fitting

All fitting methods mutate the geometry in place and return self:

from raygeo.geo import Geometry

# Simplify a path
g.simplify(tolerance=0.1)

# Convert curves to line segments
g.linearize(tolerance=0.01)

# Fit arcs and beziers to linear data
g.fit_arcs(tolerance=0.5)
g.fit_curves(tolerance=0.5, beziers=True, arcs=True)

# Convert geometry to polygons (returns list, does not mutate)
polygons = g.to_polygons(tolerance=0.01)

Self-Intersection Detection

g.has_self_intersections()          # check for self-intersections
g.intersects_with(other_geometry)   # check intersection with another geometry
g.encloses(other_geometry)          # check if this fully encloses another

Serialization

# Serialize to dict (JSON-safe)
data = g.to_dict()

# Deserialize from dict
g2 = Geometry.from_dict(data)

# Pickle support (via __reduce_ex__)
import pickle
g3 = pickle.loads(pickle.dumps(g))

Documentation

Full API reference documentation is generated from the source type stubs. Run make docs to build it locally — this produces Markdown pages in docs/api/ with inline visual examples.

The docs are also published online with the RayForge Developer Docs.

Development

Prerequisites

• Rust toolchain (latest stable)
• Python 3.10+
• maturin (pip install maturin)
• Node.js (only needed for make lint-python, which runs pyright via npx)

Quick Start

# Create and activate a virtual environment (Unix)
python -m venv .venv
source .venv/bin/activate   # on Windows: .venv\Scripts\activate

# Install build tool and build the extension
pip install maturin pytest
make dev                   # builds Rust extension and installs into venv

# Run tests
make test

# Full check (lint + test)
make check

Available Make Targets

Target	Description
make dev	Build and install into the active venv
make build	Build release wheel to dist/
make test	Run pytest
make lint	Lint Rust + Python (including pyright)
make format	Auto-format Rust + Python
make check	Lint + test
make stubs	Regenerate .pyi type stubs
make docs	Build API docs with inline visual examples
make visual	Launch Streamlit visual test playground

Visual Testing

The make visual target launches an interactive Streamlit playground with real-time plots for geometry construction, polygon booleans, curve fitting, image processing, SVG parsing, tab operations, overscan, lead- in/out, merging, rasterization, concave hull, and nesting.

pip install -e ".[visual]"
make visual

See Visual Testing for a full walkthrough of every page and its controls.

License

MIT