Splitting an R application into modular parts

toyApp.R

#
# This is a toy example of splitting an R application
# into three distinct parts:
#
#   - Loading the data
#   - Calculating the analytics
#   - Rendering an RMarkdown document
#
# Note: The RMarkdown doc appears in a companion file, toyDoc.Rmd.
#
# The split is useful because it cleanly moduarizes
# the application. We can likely change one module
# (loading, calculating, or rendering) without breaking
# another module.
#
# Also, by separating the components, we can more easily
# debug each one. For example, by splitting the (potentially)
# complicated analytics from the rendering, we can debug
# it before and outside the RMarkdown document.
#
library(rmarkdown)

loadData = function() {
  return(mtcars[,c("mpg", "disp", "wt")])
}

doAnalytics = function(data) {
  model = lm(mpg ~ disp + wt, data=data)
  list(theModel = model,
       theData = data )
}

#
# This is the application's top-level code
#
data <- loadData()
analytics <- doAnalytics(data)

# By passing the 'analytics' list as the RMarkdown environment,
# we make its contents available to R code in the doc.
rmarkdown::render("toyDoc.Rmd", envir=analytics)

toyDoc.Rmd

---
title: "Toy Report"
output: html_document
---

This study demonstrates the clear relationship in car design
between miles per gallon (*mpg*), displacement (*disp*), and
vehicle weight (*wt*).

# Linear model
```{r, echo=FALSE}
print(summary(theModel))
```

# Appendix: Sample data
```{r, results="asis", echo=FALSE}
knitr::kable(theData)
```