Upgrading to 3.0

ObjectNat 3.0 runs its computationally intensive parts in a compiled Rust engine, objectnat._rust, and adds 2.5D and grid methods on top of it. No function is removed or renamed, but the parameters of the former Python implementations are removed, some results differ and installing from source needs a Rust toolchain. Release 2.2.0 already has the new engine and still accepts the removed parameters with a warning, so this page applies to upgrades from any 2.x release.

What changes for existing code

Change	Effect	What to do
Radius isochrones and coverage follow the network	With geometry_type="radius" every reachable node is buffered by its remaining budget, but by no more than off_network_distance (default 150 m), so the area no longer grows into a straight-line circle around the start	off_network_distance=None keeps the 2.x shape
get_visibility() has one exact method	method, resolution and parallel are removed: the visibility is always exact and computed in parallel threads; max_workers limits the threads	Remove these arguments; get_visibility(points, obstacles, 100) replaces get_visibility(points, obstacles, 100, "simple")
calculate_simplified_noise_frame() always runs the visibility in parallel threads	visibility_parallel is removed	Remove the argument; visibility_max_workers limits the threads
Provision allocation runs in Rust	For a given seed the allocation differs from 2.1, but matches it statistically, and is the same on any number of cores; get_service_provision() also accepts a pandas sparse OD matrix	None needed: the model and the signature are unchanged
Noise reflections are deterministic	simulate_noise() gives the same result on every run	None needed
Installation	Wheels contain the compiled extension; building from source requires Rust; IduEdu 2.4.2 or newer	pip install -U objectnat installs a wheel, no compiler needed

What is new

• Insolation: direct sun hours on the ground (get_insolation()) and on facades floor by floor (get_facade_insolation()), among buildings and terrain, for a day or a period.

• Visibility Analysis: 2.5D viewsheds over building heights and terrain (get_viewshed()), isovist metrics in get_visibility(), isovist fields on a grid (get_isovist_grid()), catchments of many observers in 2D and 2.5D (get_visibility_catchment(), get_viewshed_catchment()) and observers along a pedestrian graph (get_observer_points()).

• Noise Simulation: simulate_noise_grid(), sound levels on a receiver grid around point, line and polygon sources with reflections and diffraction around building corners.

• Rasters: rasterize() and Raster turn grid results into matplotlib plots, GeoTIFF files and web maps.

• Speed: exact visibility and provision allocation are many times faster (provision on the example data takes 0.06 s instead of 2.6 s), and isochrones for many origins are faster with the same geometry.