Develop explicit package "guidelines" or "best practices"

[markdanese]

The R Consortium should identify key standardization practices that they want to encourage (not necessarily require). In many cases, there are already packages that support such practices. Two examples are standardization of statistical model results, and the location of data sets in functions, but there are probably many more.

1. The broom package could be held out as an example for packages to consider for standardizing the output of model results. Authors could be encouraged to write their own broom methods, or even to generate output consistent with broom. Not all models will work, of course, which is why it is a guideline.

2. The magrittr package illustrates the benefit of having a standard way of placing the data input for a function (i.e., as the first item). Again, there may be reasons not to do this. But in the absence of such reasons, it might be beneficial for the community to adopt a few selected practices that help people use data.

R could start the process by changing its own base functions to reflect whatever guidelines that are selected. (Or introducing new versions of base functions to retain backwards compatibility.)

[markdanese]

Another example that might benefit from guidelines is how to store variable labels -- handled in the haven, hmisc and sjplot packages (and probably others)

[jalazawa]

I don't know if it will the same "project" or "package", but an analyser for SonarQube can be a solution like this one analyserPython. A lot of language are already supported by SonarQube.