Nicole Sullivan Shot Logs Shiny App

[sullivannicole]

For this assignment, I created a Shiny app using data from the NBA Shot Logs from 2014-2015. In the future, I definitely want to try out Flexdashboard. Coding for Shiny is way too unwieldy and opaque and making even the slightest changes becomes a struggle when there are more than a couple tabs. I mean, I wrote the code, and I still had difficulties recalling how to layer code or properly insert new code!

[bensoltoff]

• Structuring the code in a pure Shiny app is really a pain in the ass. One of the best tools is to use the auto-indenting command in RStudio to automatically indent your code based on where it is nested inside of a function. It helps you quickly identify in a Shiny app when code is in the wrong location. Also, converting it to a Shiny flexdashboard allows you to construct it like a more traditional R Markdown document, so you don't have to segregate your server and UI code.
• In fact, I think this application lends itself very well to a flexdashboard approach given all the different tabs and elements you want to display
• So for some of the plots, converting it using ggplotly() makes sense (such as the time chart). For others, maybe not as much. Like the distance tab - hovering over the boxplots reveals numbers associated with the key elements of the boxplot, but I'm not sure I really need the precise values. You can make an interactive Shiny app but still use static visualizations for some aspects. Think carefully when designing an app like this what makes the most sense for each visualization component (i.e. each individual graph) whether to render static or dynamic.
• Too many columns in the raw data tab. It all runs off the screen. Is this necessary or useful for the app? Maybe instead of showing all columns, pick the most important ones. Or use a download handler function to allow users to download the raw data for an individual player/team.
• Being nitpicky, but on the graphs consider using scale_*_brewer() to pick the color palettes rather than ggplot2's default algorithm. This should implement more robust color palettes for distinguishing between discrete categories
• In your code you seem to reuse a custom theme() for each plot. Avoid code duplication! You could store this as a custom theme function at the beginning of server.R, then reference that in each of the plots.