Code Structure

Structure

cellular_raza consists of multiple crates working in tandem. It was designed to have clear separations between conceptual choices and implementation details. This approach allows us to have a greater amount of modularity and flexibility than regular simulation tools.

graph BT
    examples -.-> concepts
    examples --> core
    examples --> building_blocks
    benchmarks --> core
    benchmarks --> building_blocks
    benchmarks -.-> concepts
    core --> concepts
    building_blocks --> concepts

These crates act on varying levels of abstraction to yield a fully working numerical simulation. Since cellular_raza functions on different levels of abstraction, we try to indicate this in the table below.

crate	Abstraction Level	Purpose
cellular_raza	-	Bundle together functionality of all other crates.
concepts	High	Collection of (mainly) traits which need to be implemented to yield a full simulation.
core	Intermediate-High	Contains numerical solvers, storage handlers and more to actually solve a given system.
building_blocks	Intermediate	Predefined components of cell-agents and domains which can be put together to obtain a full simulation.
examples	Application	Showcases and introductions to different simulation approaches.
benchmarks	Application	Performance testing of various configurations.

flowchart LR
    High --> Intermediate --> Application

Offspring Crates & Other Contributions

During the development of cellular_raza multiple subfunctionalities which were only developed for internal use have been refactored and published as individually standing crates.

Crate	Description	Usage
approx-derive	Derive macros for AbsDiffEq and RelativeEq of the approx crate.	Testing
circ_buffer	A no_std statically allocated ring-buffer with optional serde support.	Save increments for a multistep solver.
crate2bib	Convert crates to bibligraphy entries.	Writing texts/publishing in general.
short_default	Use shorthand notation for creating structs with default values.	See cr_bayesian_optim
spatial-decomposition	Split apart 2D simulation domains such that they can be processed in parallel.	Experimental usage in CartesianCuboid2D
uniquevec	A Vec-like datastructure which only contains unique entries. It is no_std and makes no use of unsafe code.	Specify multiple StorageOption in correct priority without duplicates.

Furthermore, contributions to the following projects have been made: pyo3 rust-numpy plotters servo typst

Development

Language	Files	Lines	Blank	Comment	Code	Complexity	Bytes	Uloc
Rust	113	33418	2872	5350	25196	1297	1161637	0
Python	14	2125	272	83	1770	181	67148	0
TOML	31	650	70	12	568	0	18069	0
Markdown	20	434	97	0	337	0	16893	0
License	1	339	58	0	281	0	18092	0
SVG	3	334	0	0	334	0	28964	0
YAML	18	227	40	7	180	0	4856	0
Cuda	2	186	34	26	126	11	5381	0
HTML	3	84	0	3	81	0	4866	0
Shell	5	63	13	3	47	4	1341	0
WebGPU Shading Language	1	39	3	8	28	5	1140	0
Plain Text	6	20	0	0	20	0	304	0
Total	217	37919	3459	5492	28968	1498	1328691	0
Estimated Cost to Develop (organic) $925,997 Estimated Schedule Effort (organic) 13.36 months Estimated People Required (organic) 6.16