Open questions regarding target architecture & LATEX

[mlbach]

Some points that were left open in discussions:

Target architecture

Do we want the final product to be an R-package, maybe even submit it to CRAN?

Should be the most reproducible but may be more work during the development process.

Or do we want to provide a code repository that can be clowned and then used?

Reproducability could be achieved by using renv and maybe a fixed (portable) R-version.

In this way it is easy to set up at any organization and should be stable. However using renv might not correspond perfectly to the open source development philosophy.

latex issues

How can we make sure that latex and its packages are installed correctl on the target systems,this often leads to problems?

Maybe tinytex can be the solution, it can be installed via R.

[lukric]

These are very good points. Two remarks from my side:

• renv: Our group is trying out renv since some time and we are still running in issues regularly (e.g. packages can not be installed). Some of these issues definitely depend on our IT infrastructure where only very few CRAN mirrors are accessible. Does someone else have experiences?
• latex: a semi-solution could be that PDF output is optional. So if a user is not able to install tex or the correct packages it would not break the shiny app and the analyses. But I have no experiences with thinytex yet.

[alchalu]

Regarding the format of the final output: I unfortunately do not have any experience with golem apps yet, but the following book might be a good ressource Engineering Production-Grade Shiny Apps

[mlbach]

In our last call we noted that

• not everyone of us can install any version of R
• not everyone is allowed to use a portable version of R

Hence, we agreed

• that development should be geared towards and tested on R 4.0.2.
• we want to turn our code into a R package which includes one wrapper function around our shiny app (end product)