With large applications it is critical for maintainability
to properly structure your code using files and directories.
R comes with the library()
and source()
functions,
but its functionality is limited when it comes
to dividing your code into modules and expressing their dependencies.
To address this, Rhino uses the box R package, which allows you to modularize your code in a similar way to languages like Python and Java:
box::use( dplyr, # Import dplyr. Its functions can be used via `$`, e.g. `dplyr$filter`. shiny[reactive], # Import the `reactive()` function from shiny package. ) box::use( logic/data_validation, # Import the `logic/data_validation.R` module. )
Box modules force you to be explicit about the dependencies between your files and packages.
The graph of dependencies is visible at a glance in an app developed with box,
while the traditional approach (global.R
, library()
, source()
)
makes it easy to build an app which only the author understands.
Introduction of box to existing apps written without it
has helped to improve the code structure and find bugs.
The best place to learn about box is its official documentation. The discussion here will mainly focus on how to use box inside Rhino.
Rhino suggests the use of app/logic
and app/view
.
Rhino creates these directories by default.
Code that is independent of Shiny should be kept in app/logic
while code using or related to Shiny modules should be kept in app/view
.
This structure makes it easy to make a nested hierarchy of code with the help of box.
# app/logic/messages.R #' @export say_hello <- function(name) { paste0("Hello, ", name, "!") } #' @export say_bye <- function(name) { paste0("Goodbye, ", name, "!") }
Both say_hello()
and say_bye()
can be exported from app/logic/messages.R
.
box::use( app/logic/messages[say_bye, say_hello], ) #' @export greet <- function(name) { paste( say_hello(name), say_bye(name) ) }
Note that box::use()
allows for explicit attaching of function names from a module as shown above.
Modules can also be imported across directories; use code from app/logic
in app/view
.
# app/view/greet_module.R box::use( shiny[div, moduleServer, NS, renderText, req, textOutput], shiny.semantic[textInput], ) box::use( app/logic/greet[greet], ) #' @export ui <- function(id) { ns <- NS(id) div( textInput(ns("name"), "Name"), textOutput(ns("message")) ) } #' @export server <- function(id) { moduleServer(id, function(input, output, session) { output$message <- renderText({ req(input$name) greet(input$name) }) }) }
With explicit attaching of function names, it is clear from above that the code
uses shiny.semantic::textInput()
and not shiny::textInput()
.
# app/main.R box::use( shiny[moduleServer, NS], shiny.semantic[semanticPage], ) box::use( app/view/greet_module, ) #' @export ui <- function(id) { ns <- NS(id) semanticPage( greet_module$ui(ns("message")) ) } #' @export server <- function(id) { moduleServer(id, function(input, output, session) { greet_module$server("message") }) }
In main.R
, Shiny modules can be attached without attaching the function names.
The Shiny module functions are accessed via $
.
Some useful box features are also explained in the sections below.
Objects exported by an __init__.R
file can be imported from its parent directory.
Assume we have an app/foo/__init__.R
file with the following content:
#' @export bar <- "Hello!"
We can now import bar
as if it was defined in app/foo.R
:
box::use( app/foo[bar], )
This mechanism can be used in combination with reexports to make it easier to import multiple modules from a single directory.
A module can reexport objects imported from a different module
by applying #' @export
to a box::use()
statement.
Assume we have modules analysis_tab.R
and download_tab.R
in the app/view
directory.
We can reexport them from app/view/__init__.R
like this:
#' @export box::use( app/view/analysis_tab, app/view/download_tab, )
The following box::use()
statements are now equivalent:
box::use( app/view/analysis_tab, app/view/download_tab, ) box::use( app/view[analysis_tab, download_tab], )
To enhance the readability and maintainability of code, we suggest following the Rhino style guide.
The following issues were fixed in box v1.1.3, which is required by Rhino starting with v1.4.0. This section is left here for reference.
Prior to v1.1.3 box didn't support lazy-loaded data,
so e.g. box::use(datasets[mtcars])
wouldn't work
(see this issue).
It was possible to workaround it by using datasets::mtcars
instead.
Box allows trailing commas in box::use()
statements and code,
but prior to v1.1.3 they could cause problems in some circumstances:
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.