r4ds-rsuite

The goal of r4ds-rsuite is providing an update resistant version of the book by installing r4ds in its own environment, with its own packages, frozen at a time of your chosing.

I have cloned r4ds on June 6, 2019, and immediately added a package under the folder packages called r4ds.book.pkgs. This package contains all the packages required by r4ds so it can run independently from the global environment.

To be able to do this, we use rsuite, an application with a client (Windows, Linux and Mac), a package RSuite, and a RStudio addin. The application is open source and is available in Github.

I have been converting the most important and complex of my projects to rsuite administered form. It really is a time saver because the dependencies or packages do not break after doing a global update of R packages.

How to use this r4ds variation

1. Download and install the rsuite client in your machine.
2. Install the R package with rsuite install
3. Clone this repo
4. From the project root r4ds-rsuite, open a terminal, and run Rscript R/compile_book.R. The book will start building.
5. Run the bookdown or gitbook version of r4ds by running index.html under the folder work/r4ds/_book

You will notice that the folder deployment/libs has been populated only by the packages required by r4ds. The operation is the same for any of the operating systems. The R binaries are generated depending of the OS.

Project deployment

I like the idea behind rsuite. What I showed above is only one of the things that you can do with it. Additionally, you can:

1. Put several packages under a main umbrella project to manage all of them, including tests and builds.
2. Generate a stand-alone Python from Anaconda inside the rsuite project. This is a pretty neat idea for distributing ready-to-run applications, because if you share with other users, they don't even need to install Python; only R is needed.
3. Create a local copy of a whole remote repository, for instance CRAN, or selected packages, in you own server, Amazon instance, or your local drive. This is pretty handy when you are working in location with poor or slow internet, or no connection at all.
4. Build a distributable R application as a zip file, where your users don't need to install any packages. Unpack it and run it with R.
5. You can also deploy the R application within a Docker container as well. So, instead of sending to your users a zip file, you send them a link to download a Docker container including R itself.
6. There are few other things that rsuite does but haven't tested or explored yet. But the whole concept is pretty neat, preparing you for deployment.

References

• rsuite in Github
• rsuite website
• r4ds-rsuite