Time Series Inspector

The goal of this repository is to demonstrate how a simple shiny app can be built and deployed. The application provides an easy interface to explore timeseries. The inspector tool is embedded in the TSinspector R package and can be used by locally install the package or by running a docker container.

Install and use as R package

You can install the development version from GitHub with the following command:

# install.packages("devtools")
devtools::install_github("RobertoChiosa/timeseries-inspector")

Once installed you can run the application as follows:

library(TSinspector)
TSinspector::run_app()

Install and use through docker container

Another way to deploy the timeseries inspector App is to use Docker. In order to build the container image, you’ll need to use a Dockerfile. A Dockerfile is simply a text-based file with no file extension. In this case it was automatically created from the R console by using the following command from the R package golem:

golem::add_dockerfile()

The Dockerfile is then added to the root folder of the application and contains all the instructions to build the container. However, a more reliable way to create the Dockerfile is by using the function

golem::add_dockerfile_with_renv()

This allows to generate the Dockerfile by leveraging the renv package which allows a more reliable way to handle dependencies in the R environment and acts like the venv in python environments.

Build the container image.

To build the image run the following command into the main folder of the application.

docker build -f Dockerfile --progress=plain -t timeseries-inspector .

If you are usina a mac with M1 processo you should specify the building platform.

docker build -f Dockerfile --progress=plain -t timeseries-inspector . --platform linux/x86_64

The -t flag tags specifies a human-readable name for the final image. Since you named the image timeseries-inspector , you can refer to that image when you run a container. The . at the end of the docker build command tells Docker that it should look for the Dockerfile in the current directory.

The first time you build the image Docker downloads a lot of layers because you instructed the builder that you wanted to start from the rocker/verse:4.2.2 image. But, since you did not have it on your machine, Docker needed to download the image. Once the installation of the base image is done any further build will be much faster.

Run the container image.

Then run the application by starting the container.

docker run  --name TSinspector -p 80:80 timeseries-inspector:latest

After a few seconds, open your web browser to http://127.0.0.1/80. You should see your app running.

Push the container image to Docker Registry

Resources

• Golem
• Docker
• Shiny