Shiny and Software Engineering
Why does Shiny code feel different from any other R code written for data analysis or reporting?
I contemplated this question since I started learning Shiny, and here I will talk about the idea that I have in mind to answer this question. The title kind of gives it away, but it still deserves a more detailed explanation.
My Shiny journey so far
At that time I started my Shiny journey, I already was comfortable writing R code in both base and tidyverse “dialects”. But then, Shiny felt like a completely new thing. Building UIs was not hard because it is basically a couple of nested functions, as if the R code mimics HTML (and it kind of does because it generates divs and other elements eventually). However, I struggled with making server logic for many reasons.
Even after being able to make code that works well, I still can’t help but feel there are better ways to write Shiny code. I know there are always more tricks to discover no matter what kind of code I am writing. But this feeling is always the strongest when I am making Shiny apps. So, why is that?
We don’t do software engineering
If I were to mention a single reason why Shiny is different or more challenging, it is that Shiny is not just programming in general, but also has a lot of software development or engineering to it.
I couldn’t put my finger on it at first, but when I heard it in an episode of the Shiny Developer Series podcast, it made a lot of sense to me.
Yes, R programmers do have an idea about coding best practices. But the truth is, for the most part, we don’t have to follow them! Most analyses easily fit in a single script. Even the script itself is usually nothing but a bunch of top-level assignment statements and function calls, and if you slip in repetitions, hard-coding or other code smells, you almost always can get away with it.
On the other hand, based on my experience, Shiny gets very messy very
quickly. Try to put everything in a single app.R script and you will
easily find yourself drowning in hundreds of lines of code (assuming, of
course, you are not making a hello-world kind of app).
What I do about it
Well, the canonical advice for any coding domain is to keep building stuff, learn by doing, make mistakes and learn from them, etc. So I am currently doing this, experimenting with different coding styles, and learning as I go.
For example, I started splitting the code into separate scripts and sourcing
those scripts into the main app.R one. So far, it works fine. But I have
just heard of Shiny modules which, if I understood its general idea
correctly, contains a UI/server part of an app that can be separated from the
rest of the code and can also be connected to other parts or integrated into
the main app somehow. It is one of the things I am looking forward to
learning currently.
The whole Shiny ecosystem is growing on me, and although I explained how it is relatively challenging for a non-developer, I felt rewarded after successfully building tools. I will sure document further steps in the future.