rOpenSci | rOpenSci Software Peer Review: Still Improving

rOpenSci’s suite of packages is comprised of contributions from staff engineers and the wider R community, bringing considerable diversity of skills, expertise and experience to bear on the suite. How do we ensure that every package is held to a high standard? That’s where our software review system comes into play: packages contributed by the community undergo a transparent, constructive, non adversarial and open review process. For that process relying mostly on volunteer work, associate editors manage the incoming flow and ensure progress of submissions; authors create, submit and improve their package; reviewers, two per submission, examine the software code and user experience.

We are quite proud of our Software Peer Review system and do our best improving it over time. In this post, we summarize some of the more important recent changes to Software Peer Review at rOpenSci. The most recent information can always be found on our software review page.

A new and clearer name

Up until December, our system was called rOpenSci onboarding which was short but potentially misleading given the general understanding of “onboarding” as adding new employees or volunteers to an organization. We switched to rOpenSci Software Peer Review which shows the system draws elements from both Peer Review as known in academia, and Software Review as known in software development. The discussion about the name change started in July 2017; we’re glad to have finally done it!

Four new editors

Editorial team. From left to right and top to bottom: Brooke Anderson, Scott Chamberlain, Anna Krystalli, Lincoln Mullen, Karthik Ram, Noam Ross, Maëlle Salmon, Melina Vidoni.

Editorial team. From left to right and top to bottom: Brooke Anderson, Scott Chamberlain, Anna Krystalli, Lincoln Mullen, Karthik Ram, Noam Ross, Maëlle Salmon, Melina Vidoni.

Since our last post about updates in the system in September 2017, the editorial team has doubled in size to handle all incoming submissions whilst not overburdening editors. We are delighted to have welcomed Anna Krystalli, Lincoln Mullen, Brooke Anderson and Melina Vidoni. Thanks to them for joining us! Read more about Anna and Lincoln and about Brooke and Melina.

We wrote a book

Screenshot from our online book

Our guidance for authors, reviewers and editors used to be scattered over single Markdown files in the Software Peer Review GitHub repository, which was getting impractical as these files got bigger, and which was not giving a great reading experience to anyone. We have therefore consolidated all our guidance to a book created with bookdown. Its source is stored on GitHub. It is divided in three sections:

This blog post marks the official release of our guide, whose updates are tracked via a NEWS appendix. The next big update won’t be about packaging guidelines though: we’re working on adding a cover to the online book with Locke Creatives.

Recent changes in our standards

We have completed and improved our standards for package development.

As regards testing,

  • Our guide now clearly states conditions in which to add continuous integration for Windows CI.

  • We now do not only recommend httr but also crul and curl over RCurl for HTTP requests and suggest using the rOpenSci packages vcr and webmockr for testing.

  • We have added a rule of thumb for package coverage: “Test coverage below 75% will likely require additional tests or explanation before being sent for review.” on top of stating that it is best practice when all key functionalities are tested.

  • We now require that packages with Shiny apps use a unit-testing framework such as shinytest to test that interactive interfaces behave as expected.

As regards documentation,

We have also made these updates about package building.

  • We have made it explicit that packages should run on all major platforms: Windows, macOS, Linux. Exceptions may be granted packages that interact with system-specific functions, or wrappers for utilities that only operate on limited platforms, but authors should make every effort for cross-platform compatibility, including system-specific compilation, or containerization of external utilities.

  • Our guidance now states that argument naming and order should be consistent across functions that use similar inputs.

Of particular interest are also these changes to our policies and process.

Our policies and guidance have always been fluid, but as a result, keeping track of changing policy and package-building guidance has been challenging for authors and reviewers. With the release of developers guide, changes to requirements will be tracked in our NEWS chapter, and we will be moving to quarterly releases of updates.

Issue templates

Screenshot of issue template selection

When clicking on “new issue” in the Software Peer Review Github repository, one gets to choose between submission, pre-submission and feature request/bug report.

Screenshot of submission issue template

The first two reasons for opening an issue have their own issue template, and the last one guides people towards our repository aimed at discussion of the system. We hope it makes it easier to participate.

A landing page on our website

We used to refer to Software Peer Review’s GitHub repository when mentioning the system, which was not optimal. There is now a dedicated page on our website, at the moment explaining the motivation and principles of the system, and providing related links.

Conclusion

A new name, a bigger team, its own online book… rOpenSci Software Peer Review had an eventful year and a half! We thank all package authors and reviewers for their hard work and welcome suggestions and questions about the system.