I'm in the process of preparing a manuscript which provides ancillary material in the form of a Jupyter Notebook (*.ipynb) explaining how we implemented some key parts of our results. I'd like to link to the nbviewer rendering of our ancillary material, especially since some readers may not be familiar with how to read a *.ipynb file. For instance, 1511.08044 provides several appendices in this way, each of which can be viewed with a link of the form http://nbviewer.jupyter.org/urls/arxiv.org/src/1511.08044v1/anc/1.energies_vs_vgate.ipynb. This naively seems to depend on knowing the paper identifier that arXiv will assign the manuscript once published, however, such that there doesn't seem to be a way to provide a link that works on initial submission. Is there a best practice known for providing Jupyter Notebooks that doesn't introduce this circularity?

  • I may misunderstand you, but isn't it more practical to have your own nbviewer directory, and just mention this as a link in your manuscript in references or supplementary material? – Greg Mar 13 '17 at 5:42
  • That is definitely more practical in that it gives me more control over how the notebook is hosted and linked to, but there's some big advantages to letting arXiv host the notebook. Most notably, I have more confidence in arXiv's longevity than that of my own sites. – Chris Granade Mar 13 '17 at 9:59
  • 1
    How about a GitHub account. It is free for open source or education purposes and most probably stays around for long – Greg Mar 13 '17 at 13:04
  • 1
    I've used GitHub for most of my reproducible papers so far, and it works well in some ways. That said, it doesn't ensure that the ancillary files are versioned with the paper itself, nor does it culturally fit with how most other physicists (at least in quantum info) tend to read the arXiv, I think. Looking to use arXiv's features in a way that's easier for people reading our manuscript. – Chris Granade Mar 13 '17 at 21:25

I think you're asking the wrong question here.

For the sake of persistent archiving you should put all ancillary material at a sustainably archived website specifically designed to preserve scholarly communications material e.g. Zenodo https://zenodo.org/ or Dryad http://datadryad.org/ or Figshare https://figshare.com/

Refer-to or cite the ancillary material archived on Zenodo with the Zenodo DOI that is generated upon archival of the material. From within that ancillary material which is archived, you can link to an nbviewer prettified instance of your Jupyter nb.

Direct-linking to nbviewer from within your article is a really bad idea. In five or ten years time it's very likely that the link will be broken. Always refer to ancillary material with persistent identifiers e.g. DOIs.

  • My interpretation of the question is that the ancillary material will be part of the arXiv submission (so it will be permanently archived there and easily available from the paper's arXiv page). Instead, the question is how to provide supplementary links to nbviewer to make things easier for the reader. But I agree that linking to nbviewer will not be useful in the long run. – Anonymous Mathematician Mar 15 '17 at 16:14
  • Thanks. I didn't actually know arxiv allowed "ancillary files" until now. I still think it's a bad place for the content. The separation between code, data and 'the paper' is actually quite helpful. If an author re-uses the code - they can specifically cite the DOI of the archived code repository. If an author re-uses the data - they can specifically cite the DOI or the archived data repository. If the author cites the paper - they can cite the DOI of the paper. Putting the "ancillary material" on arxiv doesnt aid granularity of citation in my opinion. – rmounce Mar 15 '17 at 16:24
  • To clarify, I do use Zenodo and Figshare extensively in my papers. When I do so, I also provide an nbviewer link separate to the DOI since many researchers in my field (quantum info) do not know how to read, edit, or reuse *.ipynb files. Moreover, I maintain open-source libraries and DOI-citable that cover most of my work; the ancillary material is often short and explanatory rather than directly reusable. For such short ancillary files, I want to make them as readable as possible while also maintaining permanence by providing it in a way readers expect, namely by using arXiv. – Chris Granade Mar 16 '17 at 21:13

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.