HARDS: Hierarchical Arbitrary Data Storage
A pure-Python library to manage hierarchical datasets on the filesystem.
Overview
HARDS is a simple Python package to facilitate the management of hierarchical datasets where there is a parent-child relationships between data sets.
The API is modelled around three ‘objects’ of data:
- a datapoint – represents a single ‘observation’ of a system with certain inputs, outputs, and metadata
- a dataset – contains many related datapoints and may be related to other datasets such that it has access to its ‘parent’s’ datapoints
- a database – contains the top-level datasets
HARDS was originally designed for Bayesian Optimisation applications, where a naïve (Monte Carlo) sampling approach precedes intelligent algorithms. These intelligent algorithms rely on data from previous sampling campaigns, creating a parent-child relationships between the datasets.
Who it’s for
HARDS is for computational scientists running Monte Carlo simulations or working with hierarchical datasets.
Licence
MIT
Release date
29 May 2025.
Programming languages
Python