Lib.rs

Highlights

• Generic 1, 2, ..., n-dimensional arrays
• Owned arrays and array views
• Slicing, also with arbitrary step size, and negative indices to mean elements from the end of the axis.
• Views and subviews of arrays; iterators that yield subviews.

Status and Lookout

• Still iterating on and evolving the crate
  • The crate is continuously developing, and breaking changes are expected during evolution from version to version. We adopt the newest stable rust features if we need them.
• Performance:
  • Prefer higher order methods and arithmetic operations on arrays first, then iteration, and as a last priority using indexed algorithms.
  • Efficient floating point matrix multiplication even for very large matrices; can optionally use BLAS to improve it further.

Crate Feature Flags

The following crate feature flags are available. They are configured in your Cargo.toml.

• std

  • Rust standard library (enabled by default)

  • This crate can be used without the standard library by disabling the default std feature. To do so, use this in your Cargo.toml:

    [dependencies]
    ndarray = { version = "0.x.y", default-features = false }
    

  • The geomspace linspace logspace range std var var_axis and std_axis methods are only available when std is enabled.

• serde

  • Enables serialization support for serde 1.x

• rayon

  • Enables parallel iterators, parallelized methods and par_azip!.
  • Implies std

• approx

  • Implementations of traits from version 0.5 of the [approx] crate.

• blas

  • Enable transparent BLAS support for matrix multiplication. Uses blas-src for pluggable backend, which needs to be configured separately (see below).

• matrixmultiply-threading

  • Enable the threading feature in the matrixmultiply package

How to use with cargo

[dependencies]
ndarray = "0.15.0"

How to enable blas integration

Blas integration is an optional add-on. Without BLAS, ndarray uses the matrixmultiply crate for matrix multiplication for f64 and f32 arrays (and it's always enabled as a fallback since it supports matrices of arbitrary strides in both dimensions).

Depend and link to blas-src directly to pick a blas provider. Ndarray presently requires a blas provider that provides the cblas-sys interface. If further feature selection is wanted or needed then you might need to depend directly on the backend crate's source too. The backend version must be the one that blas-src also depends on.

An example configuration using system openblas is shown below. Note that only end-user projects (not libraries) should select provider:

[dependencies]
ndarray = { version = "0.15.0", features = ["blas"] }
blas-src = { version = "0.8", features = ["openblas"] }
openblas-src = { version = "0.10", features = ["cblas", "system"] }

Using system-installed dependencies can save a long time building dependencies. An example configuration using (compiled) netlib is shown below anyway:

[dependencies]
ndarray = { version = "0.15.0", features = ["blas"] }
blas-src = { version = "0.8.0", default-features = false, features = ["netlib"] }

When this is done, your program must also link to blas_src by using it or explicitly including it in your code:

extern crate blas_src;

The following versions have been verified to work together. For ndarray 0.15 or later, there is no tight coupling to the blas-src version, so version selection is more flexible.

ndarray	blas-src	openblas-src	netlib-src
0.15	0.8	0.10	0.8
0.15	0.7	0.9	0.8
0.14	0.6.1	0.9.0
0.13	0.2.0	0.6.0

Recent Changes

See RELEASES.md.