README.md

Demonstration of using R/Shiny to interactively explore a real-world, multidimensional dataset, namely, building occupancy data. How to run and view this application:

1. From within RStudio.

2. Principally, it's hosted at https://gjanee.shinyapps.io/waitz. Deployment is through RStudio. The application can also be deployed directly from GitHub, and deployment can even be automated, by creating a Docker build file (see next; the included Dockerfile is a start, but it will need to call rsconnect::setAccountInfo and rsconnect::deployApp instead of shiny::runApp), then creating a GitHub Action to build and run a container. It will be necessary to pass authentication credentials in as GitHub repository secrets. See here and here for examples.

3. As a container. Dockerfile is an example Docker build file. Build a Docker image for the application using docker built -t waitz ., run using docker run -p 8080:8080 waitz, and view at http://localhost:8080. N.B.: the build process will fail unless newer versions of R and packages, as recorded in the renv lock file here, are used.

4. Using shinylive, which obviates the need for a server and provides a means of running Shiny applications entirely within the user's browser: doesn't work, at least for this application. The application doesn't run at all in Safari, and only partially works in Firefox. Additionally, it takes a long time to load the application, and the application runs so slowly as to be unusable.

5. On Binder, but only with great difficulty. First, Dockerfile must be removed; Binder will attempt to build a container if it detects Dockerfile, but that succeeds only if the build file performs several critical, Binder-specific actions. Second, renv must be removed or at least disabled (at minimum, by removing .Rprofile) and replaced with an old-style install.R script. Then, bugs in Binder (as of this writing) prevent the application from launching by appending /shiny/waitz to the URL as stated in the Binder documentation; instead, it will be necessary to append /rstudio and then manually launch the application from within the RStudio instance hosted by Binder. Quite painful.

6. On Google Colab: not possible. A Shiny application requires that a port be exposed. Colab provides a way to map a port to the outside world from Python, but not from R.