crgutils is a Python library for loading, validating, and evaluating ASAM OpenCRG road surface files.

Supported formats and capabilities:

• ASCII .crg inputs in LRFI and LDFI formats
• Binary .crg inputs in KRBI and KDBI formats
• Road-space queries (u, v) → z, (u, v) → (x, y), (u, v) → (phi, kappa)
• World-space queries (x, y) → z, (x, y) → (u, v), (x, y) → (phi, kappa)
• All five border modes: NONE, EX_ZERO, EX_KEEP, REPEAT, REFLECT
• Header-defined evaluation options and pending runtime modifiers

Core package:

uv sync

Core package with test dependencies:

uv sync --extra dev

If you also want the demo workspace members, especially the visualizer's extra dependencies:

uv sync --all-packages

Python 3.11+ is required.

import crgutils

ds = crgutils.read("samples/crg-txt/handmade_straight.crg")  # load + validate + apply modifiers
cp = ds.create_contact_point()                                 # default options

z = cp.eval_uv_to_z(10.0, 0.0)            # elevation at road (u=10, v=0)
x, y = cp.eval_uv_to_xy(10.0, 0.5)        # road → world coords
u, v = cp.eval_xy_to_uv(x, y)             # world → road coords
phi, kappa = cp.eval_uv_to_pk(10.0, 0.0)  # heading and curvature

Override evaluation options when creating the contact point:

cp = ds.create_contact_point(border_mode_u=crgutils.BorderMode.REPEAT)

Exported types:

cp.eval_uv_to_z(u, v)       → float            # elevation from road coords
cp.eval_xy_to_z(x, y)       → float            # elevation from world coords
cp.eval_uv_to_xy(u, v)      → (float, float)   # road → world
cp.eval_xy_to_uv(x, y)      → (float, float)   # world → road
cp.eval_uv_to_pk(u, v)      → (float, float)   # heading, curvature
cp.eval_xy_to_pk(x, y)      → (float, float)   # heading, curvature
cp.eval_uv_to_z_grid(u, v)  → np.ndarray       # vectorised elevation
cp.eval_xy_to_z_grid(x, y)  → np.ndarray       # vectorised elevation
cp.with_options(**overrides) → ContactPoint     # new context, shared dataset

Bundled sample files live in:

samples/crg-txt/    ASCII CRG files (straight, curved, banked, sloped, circle, …)
samples/crg-bin/    Binary CRG files (Belgian block cobblestone, country road, …)

Both the demo scripts and the test suite load from these directories.

The repository has two demo areas:

• demo/c-api/ contains Python ports of the upstream OpenCRG C API demos
• demo/viz/ contains interactive visualization tools with their own dependencies

uv run python demo/c-api/simple.py                        # xy→z on a 2D grid
uv run python demo/c-api/reader.py                        # header / channel / road info
uv run python demo/c-api/curvature.py                     # curvature profile
uv run python demo/c-api/eval_z.py                        # uv→z, round-trip z, pk
uv run python demo/c-api/eval_xy_uv.py                    # uv→xy→uv round-trips
uv run python demo/c-api/eval_options.py                  # 29 option/modifier tests

Each script defaults to a file in samples/crg-txt/ and accepts an explicit path:

uv run python demo/c-api/reader.py samples/crg-bin/belgian_block.crg
uv run python demo/c-api/eval_xy_uv.py samples/crg-txt/handmade_curved.crg
uv run python demo/c-api/eval_options.py -t 5   # run only test 5

Pass --help to any script for usage details.

The visualizer lives in the viz workspace member and includes both a marimo app and a notebook:

uv sync --package viz
uv run --package viz marimo edit demo/viz/crg_viz.py
uv run --package viz jupyter notebook demo/viz/render.ipynb

The marimo app lets you browse bundled sample files or point it at any local .crg file.

uv run pytest

The test suite covers the loader, all evaluation functions, all five border modes, modifiers, check(), ContactPoint option inheritance, and round-trip accuracy.

src/crgutils/   library source
samples/        bundled CRG sample files (crg-txt/ and crg-bin/)
demo/c-api/     Python ports of the OpenCRG C API demos
demo/viz/       visualization app and notebook
tests/          pytest test suite

• The binary Belgian-block sample contains all-NaN edge channels that produce expected RuntimeWarnings from NumPy when computing array summaries; the file still loads and is fully covered by tests.