“Why is your Shiny app slow?”

So your Shiny app is slow? Well, I hate to break it to you, but it’s probably because of one, or both, of these reasons.

  1. Inefficient code being run in the app (probably your code, BTW)
  2. Too many calculations/processing taking place in the app

I recently went through this after making some changes to my weather_dashboard pet project. They were poorly thought through, made the dashboard slow and were generally a bad idea in the 1st place. However, while doing the work to undo them, I thought it would be a good to show HOW someone might go about addressing a similar situation. After all, this is a common scenario - a dashboard that’s been around for awhile, had continuous *improvements* over time and now has become slow. What I think is important in solving this sort of thing aren’t the tools, but the philosophical approach that’s used.

Focus on the right things!

Often when slow dashboards are brought up, the intent is to find some magic runtime settings in the application framework (Shiny in this case), or server tweaks, which will gloss over the real problems. While this sometimes works out, generally there are better solutions - if we actually know what is going on in the app!

Step 1: Understand what your application is doing

Seems like a no-brainer, right? So for example, my weather_dashboard app does the following…

flowchart LR
  A["getData()"] --> B["cleanData()"]
  B --> C{filtered_by_city}
  C --> D("createPlot()")
  C --> E("createTable()")
Figure 1

…surprisingly though, people often can’t seem to do this with their Shiny app. I think this is usually because of some frankly bad coding practices wherein people jam everything inside their shiny app as in-line code. So, if you find yourself in this boat, break your inline code into functions. This will make the app easier to maintain AND easier to profile, which we’ll talk about next.

Step 2: Profile your app

Once you know WHAT your app is doing, the next step in figuring out where it’s spending the time.