KBI0021: Publishing a DataLad dataset to Zenodo#


Michał Szczepanik <m.szczepanik@fz-juelich.de>




zenodo, special remote



This KBI discusses the currently available methods for publishing a DataLad dataset to Zenodo.

Partial solution 1: archive just the Git repository#

An easy way to index a dataset in Zenodo is by publishing it to GitHub:

  • publish the dataset to GitHub,

  • test whether a clone made from GitHub is able to retrieve data (from other locations),

  • log in to Zenodo and add the GitHub repository,

  • create a release on GitHub to trigger DOI assignment on Zenodo.

Detailed instructions are available from the Canadian Open Neuroscience Platform.

This method can be used to obtain a Zenodo record that corresponds to a versioned dataset. However, as GitHub can only hold the Git part of the dataset, no annexed files will be stored in Zenodo. Annexed file contents will need to be published to a special remote elsewhere.

Partial solution 2: upload files using an existing special remote implementation#

We are aware of one git-annex special remote implementation for Zenodo, git-annex-remote-zenodo. However, it is described as a work in progress, and in our testing it showed limited functionality (with a small patch we were able to push files to Zenodo, and retrieve them back, but only in one and the same dataset – not in a dataset clone).

See the discussion in the DataLad Handbook repository issue #941 for a description of required steps (including changes to code).