Ready4R (2024/04/02): At a Crossroads

Hi Ted,

I would like to learn about unit/automated tests while writing functions/putting together scripts that perform a series of tasks. I am not a beginner but I have not been able to find decent resources online about that ("unit tests from scratch") ; every one that I found assumes you already know what it is and how to use it.

This probably ties to general guidelines about writing packages. What I found is that there are great resources out there on how to write a package from scratch (from Jenny Bryan et al.), but I had to figure out on my own what to do when I have my own script (that does a series of things in a linear fashion) that I want to turn into a package (which to be honest is probably the most common case among R users - I think there are very few entry-level users who start a project/task/analysis from writing a package).

Somehow the conceptual leap that a package is a series of functions so you have to break down your script into a series of functions was not explicit in online resources, and even after one figures this one out, how do you deal with the fact that your script is a linear progression of analytical steps but a package is a collection of independent functions?

So I think a short explanation/demo/tutorial on: "here is my R script that does ABC, and here is how to turn it into a package" would be very helpful.

And thank you for the great resource so far, keep up the good work :-)

Hey Ted!

Thanks for the newsletter, been loving it so far. I'm interested in any tips and workflows with dealing with big data sets. I know there's Arrow and DuckDB but would love to hear if you have any recs or workflows. I am currently dealing with large data bases and I need to pull data from them but I ran out of memory most of the time. I'm not a data scientist so I struggle more than one, I suppose. Thanks!!

Hi,

Thanks for the tutorials. I believe those who want to establish a solid base on statistics and ML, would be happy to see your tutorials on ML data pipeline (from creating a project, use virtual environment for package management, develop ML and deploy the model using Shiny, bslib libraries). The model will not only generate predictions but will also create some key metrics in graphical format, like AUC_ROC curve, precision recall curve or feature importance. I know it's too much but this type of workflow is indispensable now-a-days.

Thanks