Merge pull request #2 from warlock-labs/feat/basic-operations

.github/workflows/CI.yml

@@ -3,13 +3,13 @@ name: CI
 on:
   push:
     branches:
-      - master
+      - main
     tags:
       - 'v*.*.*'
   pull_request:
     types: [ opened, synchronize, reopened ]
     branches:
-      - master
+      - main
 env:
   CARGO_TERM_COLOR: always
 

Cargo.toml

@@ -8,13 +8,13 @@ license = "MIT"
 readme = "README.md"
 repository = "https://github.com/warlock-labs/dimensionals"
 name = "dimensionals"
-version = "0.1.1"
+version = "0.2.1"
 edition = "2021"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-num = "0.4.3"
+num-traits = "0.2.19"
 
 [lib]
 
@@ -23,4 +23,4 @@ criterion = "0.5.1"
 
 [[bench]]
 name = "benchmarks"
-harness = false
+harness = false

README.md

@@ -5,7 +5,9 @@
 
 # dimensionals
 
-Dimensionals is a Rust library for working with n-dimensional data. It provides a flexible and efficient multidimensional array implementation with a generic storage backend.
+Dimensionals is a Rust library for working with n-dimensional data. It provides
+a flexible and efficient multidimensional array implementation with a generic
+storage backend.
 
 ## Motivations
 
@@ -18,12 +20,10 @@ The key motivations behind Dimensionals are:
 
 ## Features
 
-- Generic over element type and number of dimensions
-- Efficient storage using a linear memory layout
-- Index and mutable index operations
-- Arithmetic operations for 1D and 2D arrays
-- Convenient macros for array creation
-- Extensible with custom storage backends
+- Generic over an element type, number of dimensions, and storage backend
+- Iterators, slices, indexing, and other standard Rust traits
+- Ergonomic and idiomatic `std::ops` implementations for arithmetic operations
+- Convenient macros for vector and matrix creation
 
 ## Usage
 
@@ -55,6 +55,7 @@ For more examples and usage details, see the [API documentation](https://docs.rs
 
 The following features and improvements are planned for future releases:
 
+- Arithmetic operations for 1D and 2D arrays
 - SIMD support for improved performance on CPU.
 - GPU support for offloading computations to compatible GPUs.
 - Comprehensive scalar, vector, matrix, and tensor algebra operations.
@@ -64,21 +65,24 @@ The following features and improvements are planned for future releases:
 
 ## Performance
 
-The `LinearArrayStorage` backend stores elements in a contiguous `Vec<T>` and computes element indices on the fly. This provides good cache locality for traversals, but may not be optimal for sparse or very high dimensional arrays.
+The `LinearArrayStorage` backend stores elements in a contiguous `Vec<T>` and computes element indices on the fly. This
+provides good cache locality for traversals, but may not be optimal for sparse or very high dimensional arrays.
 
 Alternative storage backends can be implemented by defining a type that implements the `DimensionalStorage` trait.
 
 ## Contributing
 
-Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests on the [GitHub repository](https://github.com/warlock-labs/dimensionals).
+Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests on
+the [GitHub repository](https://github.com/warlock-labs/dimensionals).
 
 ## License
 
 This project is licensed under the [MIT License](https://choosealicense.com/licenses/mit/).
 
 ## Acknowledgements
 
-This project is inspired by and builds upon ideas from several existing multidimensional array libraries in Rust and other languages.
+This project is inspired by and builds upon ideas from several existing multidimensional array libraries in Rust and
+other languages.
 
 ## Contact
 

TODO.md

@@ -0,0 +1,65 @@
+- [ ] Use safe rust in iterators
+- [ * ] Use safe rust in indexing
+- [ ] Add tensor macro for creating tensors
+- [ ] Remove the need for phantom data markers
+- [ ] Move shape data to type-system such that it is known at compile time
+- [ * ] Support common arithmetic operations
+- [ * ] Use safe rust in arithmetic operations
+- [ ] Support reshaping
+- [ ] Support appending
+- [ ] Support removing
+- [ ] Linear algebra functions
+- [ ] Support for common statistical functions
+- [ ] Support for geometric functions like Brownian motion
+- [ ] Support for GPU offloading
+- [ ] Support for SIMD
+- [ ] Matrix multiplication
+- [ ] Support Apache Arrow or safetensors storage backend?
+- [ ] Support Pola.rs integration
+- [ ] Support plotly-rs integration
+- [ ] Support argmin-rs integration
+- [ ] Support for rayon
+- [ ] Feature flags for enabling/disabling features
+- [ ] no-std
+- [ ] Support for WebAssembly
+- [ ] Support for WebGPU
+- [ ] Support for SVM target
+
+
+The goal is for this library to have an intuitive interface for working with
+n dimensional/tensor data, while interfacing with libraries like Polars, and
+by proxy plotly-rs. It should have linear algebra functionality, and integration
+with argmin-rs for optimization. It should have common statistical functions
+such that it is able to work in a statistical computing environment.
+
+The signature for creating a Dimensional is ugly in that N must be speficied twice, a generalized 
+builder pattern, or obviation of the need to store twice would be ideal.
+
+In lib.rs:
+
+The TODO for the tensor macro is still present. This is not a bug, but a reminder for future implementation.
+
+
+In core.rs:
+
+The PhantomData<T> in the Dimensional struct is still present but unused. You might consider removing it if it's not needed for type invariance.
+
+
+In iterators.rs:
+
+The mutable iterator still uses unsafe code. While this is not necessarily a bug, it's worth noting that it introduces potential safety risks if not handled carefully.
+
+
+In storage.rs:
+
+No significant issues found. The implementation looks correct and well-tested.
+
+
+
+Overall, the code appears to be functioning as intended. The main points to consider are:
+
+Removing unused PhantomData if it's not needed.
+Potentially finding a safe alternative to the unsafe code in the mutable iterator, if possible.
+Implementing the tensor macro in the future, as noted in the TODO.
+
+These are not critical issues, but rather areas for potential future improvement. The library as it stands should work correctly for its intended purpose.

benches/benchmarks.rs

@@ -1,6 +1,9 @@
 use criterion::{black_box, criterion_group, criterion_main, Criterion};
 use dimensionals::{Dimensional, LinearArrayStorage};
 
+// TODO: This needs meaningful benchmarks for common operations useful in
+// quantitive situations
+
 fn bench_dimensional_array_creation_zeros(c: &mut Criterion) {
     let shape = [1000, 1000];
     c.bench_function("dimensional_array_creation_zeros", |b| {