ndarray-c-nim Documentation

Nim bindings for the ndarray-c library - a numpy-like ndarray library for C with multi-dimensional arrays, OpenMP parallelization, and BLAS-optimized operations.

Repository: https://github.com/jailop/ndarray-nim

Overview

ndarray-c-nim provides a clean, idiomatic Nim interface to the ndarray-c library with automatic memory management through Nim's destructors and move semantics.

Features

Multi-dimensional arrays (ndim ≥ 2)

OpenMP parallelization for performance

BLAS-optimized operations for linear algebra

Automatic memory management - no cleanup needed!

Simple API - use @[2, 3] instead of @[2.csize_t, 3]

Type-safe Nim bindings

Quick Start

import ndarray

# Create a 2x3 array of ones
let arr = newOnes(@[2, 3])

# Set value at position (1, 2)
arr.set(@[1, 2], 42.0)

# Print the array
arr.print("My Array", 2)

# No cleanup needed - automatic memory management!

Documentation

Module Reference - Complete API documentation with all available functions, types, and examples
Index - Alphabetical index of all symbols

Installation

Prerequisites

First, install the ndarray-c library:

git clone https://github.com/jailop/ndarray-c.git
cd ndarray-c
mkdir build && cd build
cmake ..
make
sudo make install

Install Nim bindings

nimble install ndarray_c_nim

Key API Categories

Array Creation

newZeros, newOnes, newFull
newFromData
newRandomUniform, newRandomNormal
newArange, newLinspace

Element Access

get(pos), set(pos, value)
setSlice, fillSlice, getSlicePtr, copySlice

Arithmetic Operations (In-place)

add, mul - element-wise operations
addScalar, mulScalar - scalar operations
axpby - linear combination
clipMin, clipMax, clip - clipping
abs, sign - element-wise functions

Comparison and Logical (Returns new array)

newEqual, newLess, newGreater
newEqualScalar, newLessScalar, newGreaterScalar
newLogicalAnd, newLogicalOr, newLogicalNot
newWhere - conditional selection

Linear Algebra (Returns new array)

newMatmul - matrix multiplication
newTensordot - tensor contraction
newTranspose - transpose

Array Manipulation

reshape - reshape in-place
newTake - extract slice
newStack - stack arrays along new axis
newConcat - concatenate arrays

Aggregation

newAggregate(axis, type) - aggregate along axis
scalarAggregate(type) - aggregate all elements
Types: aggrSum, aggrMean, aggrStd, aggrMax, aggrMin

I/O Operations

save(filename) - save to binary file
newLoad(filename) - load from binary file

Examples

Matrix Multiplication

import ndarray

let a = newOnes(@[2, 3])
let b = newOnes(@[3, 4])
let c = a.newMatmul(b)

c.print("Result", 2)

Aggregation

import ndarray

let arr = newArange(@[3, 4], 0.0, 12.0, 1.0)

# Sum along axis 0
let sumAxis0 = arr.newAggregate(0, aggrSum)

# Mean of all elements
let meanAll = arr.scalarAggregate(aggrMean)
echo "Mean: ", meanAll

Conditional Operations

import ndarray

let data = newArange(@[2, 3], 0.0, 6.0, 1.0)
let mask = data.newGreaterScalar(2.5)
let zeros = newZeros(@[2, 3])
let filtered = newWhere(mask, data, zeros)

filtered.print("Filtered", 2)

Important Notes

Minimum dimension: All arrays must have ndim ≥ 2
Type casting: The library works with cdouble (C double precision floats)
Memory: Arrays use C heap allocation, managed by Nim destructors
Simple syntax: Use @[2, 3] instead of @[2.csize_t, 3] - both work!

Links

Generated with Nim's documentation generator