B-E-+
\ \ \
+-A-D
Reproducing research results requires knowing exactly which version of the data was used and understanding the steps taken to clean and analyse it. This can be challenging in distributed environments with many contributors.
bead is a simple command-line tool and archive format that
systematically documents data work]ows. Each saved bead captures a single processing step: the
input data (via unique references), the code used, and the resulting outputs—whether cleaned datasets,
visualisations, or reports. By recording these steps and linking them via input references, bead facilitates
easy review, maintenance, and re-execution of computations, while also fostering collaboration.
Download from https://github.com/ceumicrodata/bead/releases
Given a discrete computation of the form
output = function(*inputs)
a BEAD captures all three named parts:
output - data files (results of the computation)function - source code files, that when run hopefully compute output from inputsinputs - are other BEADs’ output and thus stored as references to those BEADsAs a special case pure data can be thought of as constant computation having only output but neither inputs nor source code.
A BEAD has some other metadata - notably it has a kind property which is shared by
different versions of the conceptually same computation (input or function may be updated/improved)
and a timestamp when the computation was frozen.
The kind and timestamp properties enable a meaningful update operation on inputs.
New computations get a new, universally unique kind (technically an uuid).
Although most of the important stuff is implemented, there are still some raw edges.
Documentation for the tool is mostly the command line help.
The doc directory has concept descriptions, maybe some use cases,
but there are also design fragments - you might be mislead by them as they
are nor describing the current situations nor are they showing the future.
FIXME: clean up documentation.
Ensure you have Python 3.6+ installed.
Run make executables to create the bead tool:
$ python build.py
This generates one-file executables for unix, mac, and windows in the executables directory:
bead unix & macbead.cmd windowsMove/copy the bead binary for your platform to some directory on your PATH.
E.g.
$ cp executables/bead ~/.local/bin
If you test it, please give feedback on
Any other nuisance reported - however minor you think it be - is important and welcome!
Thank you for your interest!
Updated by script (./ci), text in this section and afterwards will be overwritten