Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Quick and Dirty (QnD): Python Binary Storage Interface

This very simple end-user-friendly API for storing numpy ndarrays in binary files is written as a frontend capable of supporting multiple backend file formats. Current backends include an h5py wrapper for HDF5 files and pure python implementations of the PDB and netCDF3 file formats. The qnd API concentrates on declaring, writing, querying, and reading numpy ndarrays (excepting dtype 'O' arrays of arbitrary python objects). Additionally, qnd maps python dicts (with str-valued keys) to HDF5 groups or PDB directories, and defines its own mapping from python lists to groups/directories with automatically generated member names of the form _0, _1, _2, etc. The support for both dict and list objects means that qnd easily handles any data structures that can be mapped to the JSON data interchange format, which is pretty nearly anything.

Another qnd innovation is a simplified interface for storing the history of changing arrays. In HDF5 or PDB or netCDF3, these are presented as arrays with an unlimited dimension, but are in fact stored in discontiguous chunks or blocks in all three cases. The qnd API, by contrast, introduces a recording mode in which writing to a variable creates its next chunk rather than overwriting its previous value. A complementary goto mode allows you to retrieve chunks in the corresponding way.

Not all backends support all of the API features; for example, the netCDF3 backend cannot handle groups or lists beyond the root group of the entire file and the lists qnd associates with record variables.

Although the qnd interface provides a very general framework for binary storage, informed by the excellent netCDF data model, in detail qnd is highly pythonic. As much as possible, qnd file, group, or list handles behave like python dicts or lists, with the same operators, iterators, and important method names as the workhorse python and numpy objects they represent.


QnD was created by David H. Munro.


QnD is distributed under the terms of the BSD-3 License.

All new contributions must be made under this license.

See LICENSE and NOTICE for details.

SPDX-License-Identifier: BSD-3-Clause

Lawrence Livermore tracking ID: LLNL-CODE-807802


Quick n Dirty Python package for working with binary files








No releases published


No packages published