Open Source/Scientific Computing Library for DartNetherlands

Numd

An open-source numerical computing library that brings NumPy-style n-dimensional arrays to Flutter and Dart. Native C++ backend using xtensor and templates, with a clean Dart API over FFI. Built with QHarbor Research in the Netherlands and used in their scientific research workflows.

DartC++xtensorHDF5Open Source

Visit Live Site

Headline

Open-source scientific library used in research workflows

01 The brief

What they needed.

Dart had no NumPy equivalent, which kept scientific Dart and Flutter workflows confined to single-precision floats and made the language an awkward choice for research pipelines. QHarbor Research, a Netherlands-based scientific group, needed a library that could handle integer arrays, complex numbers, mixed-type arithmetic with NumPy-style promotion rules, and HDF5-backed datasets, all with a clean public API and the performance ceiling of native C++.

02 Our approach

How we built it.

We built the core on a native C++ backend using xtensor and template-based architecture, exposed to Dart through a clean FFI bridge. The result is NumPy-style n-dimensional arrays with native performance and a Dart-ergonomic API.

The library ships with six data types: float32, float64, complex64, complex128, int32, and int64, with NumPy-compatible automatic type promotion across mixed-type arithmetic.

Complex number support is first-class. The Complex class covers addition, subtraction, multiplication, division, conjugation, magnitude, and phase, with automatic conversion between real and complex types when operations require it.

We contributed HDF5 integration for reading and writing large scientific datasets, data-visualisation workflows used by researchers, and unit tests, documentation, and usage examples to support open-source adoption.

03 Outcomes