code and data, Binder improvements, reproducibility

[Philip Durbin]

Hello Dataverse and reproducible research enthusiasts!

If you're not familiar with Binder, here's how we describe it in the guides: "Researchers can launch Jupyter Notebooks, RStudio, and other computational environments by entering the DOI of a dataset" https://guides.dataverse.org/en/5.13/admin/integrations.html#binder

I'm happy to report that if you have enabled the "Binder" button in your installation of Dataverse or if you simply go directly to https://mybinder.org and enter the DOI for your dataset there, tabular data files that Binder downloads from Dataverse are now* downloaded in their original format (instead of the .tab archival format).

I hope this makes it easier for you to deposit code in Dataverse that operates on the same file name and format both locally on your laptop and in Binder. How can code that expects foo.dta (Stata) expect to operate on foo.tab? This was my bad for not prioritizing this in the original implementation!

On a related note, I mentioned this in passing during a community call but early in the year we enabled the Binder button on Harvard Dataverse. So far so good. You can see some screenshots at https://github.com/IQSS/dataverse.harvard.edu/issues/208

Have fun reproducing research with Binder!

Phil

* Details about the change to repo2docker: https://github.com/IQSS/dataverse/pull/9483

p.s. I'd like to give a shout out to Craig Willis and Kacper Kowalik for hosting a service that allows the Binder button to work. Eventually, we'll improve the external tool framework to allow the path of a tool URL to be populated (instead of just query parameters): https://github.com/IQSS/dataverse/issues/9345

--

Philip Durbin
Software Developer for http://dataverse.org
http://www.iq.harvard.edu/people/philip-durbin