I’ll start with setting up a basic multi-window app using the visNetwork package, which has various .js and .css dependencies.
Load Libraries and Specify Our Inital App
As with any mwshiny application, we start by loading our libraries that we want globally accessed.
## Loading required package: shiny
Next, I’ll just naively specify a simple mwshiny app without any dependencies. My app is going to have two windows: “controller” and “visualization” windows. The controller specifies the number of nodes and edges in my graph, and the visualization window just shows our graph.
# the titles of our windows
win_titles <- c("Controller", "Graph")
ui_win <- list()
# first we add what we want to see in the controller to the list
ui_win[[1]] <- fluidPage(
titlePanel("visNetwork Dependency Example: Controller"),
sidebarLayout(
# where all our choices will go
sidebarPanel(
# choose the number of nodes, from 3 to 10
numericInput(inputId = "num_nodes",
label = "How many nodes would you like to see?",
value = 3,
min = 3,
max = 10),
# choose the number of edges, from 2 to 9
numericInput(inputId = "num_edges",
label = "How many edges would you like to see?",
value = 2,
min = 2,
max = 9),
# only build our graph when this button is clicked
actionButton(inputId = "go",
label = "Graph!")
),
# empty main panel
mainPanel()
)
)
# next we add what we want to see in the visualization window to the ui list
ui_win[[2]] <- fluidPage(
titlePanel("visNetwork Dependency Example: Graph"),
visNetworkOutput(outputId="network",
height = "800px")
)
# now we do some calculations based on our inputs, to pass to our final rendering of our network
serv_calc <- list()
serv_calc[[1]] <- function(calc, sess){
observeEvent(calc$go, {
# create our nodes data frame for the visNetowrk render plot to use
calc[["nodes"]] <- data.frame(id = 1:calc$num_nodes)
# create our edges data frame, specifying a random set of nodes to connect
calc[["edges"]] <- data.frame(from = sample(calc$num_edges, calc$num_edges, replace = T),
to = sample(calc$num_edges, calc$num_edges, replace = T))
})
}
# now we render our output!
serv_out <- list()
serv_out[["network"]] <- function(calc, sess){
# this function renders our graph, for shiny purposes
renderVisNetwork({
# don't render anything if we haven't pressed the graph! button yet!
if (!is.null(calc$nodes) & !is.null(calc$edges)){
visNetwork(calc$nodes, calc$edges)
}
})
}
Now that we have our app, let’s run it!
When we do run it, we initially see what we would expect – our controller and our visualization window, with no graph, since we haven’t pressed anything yet.
However, when we do press our button, we see… nothing. No plot in the visualization window. What’s going on?
Figuring Out Dependencies by Making a Minimal Shiny Application
So, there’s clearly something going wrong here, since our code is correct. As it turns out, the problem is that we haven’t specified the dependencies necessary to actually run our application. Let’s figure out what’s missing by using the developer console in our broswer (Ctrl/Cmd+Shift+C in Firefox or Chrome). Opening this on the visualization window, we see the HTML that is dictating our page:
If we click to open the “head” tag, we see a lot of .js and .css scripts that our page is using:
Keep those in mind. So, since our mwshiny app isn’t working, we go to the next logical step: making a minimal Shiny application that represents our visualization and checking on that rendered HTML. Since our visualization is the thing that isn’t working, we’ll just focus on making a visNetwork graph, with no interactions:
# shiny is already loaded by mwshiny
# our UI, which is going to just show a default graph
ui <- fluidPage(
visNetworkOutput(outputId="network",
height = "800px")
)
# our server function, which renders our graph
server <- function(input, output, session) {
# the necessary components of a graph with 3 nodes and 2 edges
nodes <- data.frame(id = 1:3)
edges <- data.frame(from = c(1,2), to = c(1,3))
# render our network for the UI
output$network <- renderVisNetwork({
visNetwork(nodes, edges)
})
}
# run this app!
shinyApp(ui, server)
When we run our Shiny app, we see:
Exactly what we would expect! A graph with 3 nodes and 2 edges. So, what is the difference between our Shiny app and our mwshiny app? Let’s check out the developer console on this page. When we open the head tag here, we can see that there are a few more scripts and stylesheets listed than in our mwshiny app:
Comparing this list to the one given by mwshiny, it looks like there are 4 scripts not reflected in the mwshiny app: htmlwidgets.js, vis.css, vis.min.js, and visNetwork.js. How do we import these to our mwshiny app? First, we need to figure out which packages correspond to these scripts. Luckily, there’s a simple base function that lists the scripts in each package.
Let’s start by trying to find htmlwidgets.js. Looking at the file extension, we see that this might come from the htmlwidgets package, so we can check this with a simple command:
## [1] "www/htmlwidgets.js"
There it is! So we know that we’ll have to state htmlwidgets as a dependency for mwshiny. Following a similar thought process, let’s look at the .js and .css files in the visNetwork package (there is no vis package, but it’s easy to guess where it comes from).
## [1] "doc/js/bootstrap.js"
## [2] "doc/js/bootstrap.min.js"
## [3] "doc/js/docs.min.js"
## [4] "doc/js/googleAnalytics.js"
## [5] "doc/js/ie-emulation-modes-warning.js"
## [6] "doc/js/ie10-viewport-bug-workaround.js"
## [7] "doc/js/jquery.highlight.js"
## [8] "doc/js/jquery.min.js"
## [9] "doc/js/jquery.url.min.js"
## [10] "doc/js/main.js"
## [11] "doc/js/prettify/lang-apollo.js"
## [12] "doc/js/prettify/lang-css.js"
## [13] "doc/js/prettify/lang-hs.js"
## [14] "doc/js/prettify/lang-lisp.js"
## [15] "doc/js/prettify/lang-lua.js"
## [16] "doc/js/prettify/lang-ml.js"
## [17] "doc/js/prettify/lang-proto.js"
## [18] "doc/js/prettify/lang-scala.js"
## [19] "doc/js/prettify/lang-sql.js"
## [20] "doc/js/prettify/lang-vb.js"
## [21] "doc/js/prettify/lang-vhdl.js"
## [22] "doc/js/prettify/lang-wiki.js"
## [23] "doc/js/prettify/lang-yaml.js"
## [24] "doc/js/prettify/prettify.js"
## [25] "doc/js/smooth-scroll.min.js"
## [26] "doc/js/tipuesearch.config.js"
## [27] "doc/js/tipuesearch.js"
## [28] "doc/js/tipuesearch.min.js"
## [29] "doc/js/toggleTable.js"
## [30] "htmlwidgets/lib/export/Blob/Blob.js"
## [31] "htmlwidgets/lib/export/FileSaver/FileSaver.min.js"
## [32] "htmlwidgets/lib/export/canvas-toBlob/canvas-toBlob.js"
## [33] "htmlwidgets/lib/export/html2canvas/html2canvas.js"
## [34] "htmlwidgets/lib/export/jsPDF/jspdf.debug.js"
## [35] "htmlwidgets/lib/vis/vis.min.js"
## [36] "htmlwidgets/visNetwork.js"
##
##
## [1] "doc/css/bootstrap-theme.css"
## [2] "doc/css/bootstrap-theme.min.css"
## [3] "doc/css/bootstrap.css"
## [4] "doc/css/bootstrap.min.css"
## [5] "doc/css/carousel.css"
## [6] "doc/css/prettify.css"
## [7] "doc/css/style.css"
## [8] "doc/css/tipuesearch.css"
## [9] "doc/js/prettify/prettify.css"
## [10] "html/R.css"
## [11] "htmlwidgets/lib/css/dataManipulation.css"
## [12] "htmlwidgets/lib/font-awesome/css/font-awesome.css"
## [13] "htmlwidgets/lib/font-awesome/css/font-awesome.min.css"
## [14] "htmlwidgets/lib/ionicons/css/ionicons.min.css"
## [15] "htmlwidgets/lib/vis/vis.css"
Looking through, we can see our final missing .js and .css files, so there we have our final dependency! Let’s now bring them in and run our mwshiny app accordingly.
Adding Dependencies with All Scripts to mwsApp()
So now we have htmlwidgets and visNetwork as packages we want to state as dependencies. mwsApp() has the option, depend, a named list with the names corresponding to the packages for which .js and .css files will be imported. If value for that package is NA, all .js and .css files will be imported. Otherwise, value must be a vector of strings corresponding to the .js and .css files to be imported, with location relative to the package folder.
We’re going to start by importing all of the .js and .css scripts in our dependencies:
Great! We’ve added our imported all the files, so everything should work out now, right? Well, not quite. Looking at our page, we see quite a mess:
All the pages have collapsed into one page! It seems that by importing all of the .js and .css files from our packages, we’ve created some sort of conflict. We can see above that there are a lot of opportunities with the visNetwork package, since it has a lot of scripts.
However, we know from our examination of the console that we don’t need all of those scripts – just 4. So we can specify those directly. (Note: it does work for some packages to use the simple way out and import all packages, but others, such as visNetwork, need more specificity.)
Adding Dependencies with Specific Scripts to mwsApp()
In order to add only specific dependencies, we need to find out their paths relative to the package folder. Since we’ve printed all the locations of our folders above, they should be easy enough to find and copy-paste appropriately. We’re going to put the extensions of our .js and .css files for each package in a vector of strings as the value for our depend list.
Note that order matters – these scripts will be loaded in the order that they’re specified. So, in the following, “htmlwidgets.js”is added first, then “vis.css”, and so on. If scripts later in the order depend on previous scripts, it’s important to specify the dependencies first.
Now that we’ve limited our imported dependencies to just the missing scripts, we can see that our mwshiny app works perfectly:
Thus, For JS and CSS-dependent packages, we can bring in the necessary JS and CSS scripts we need to create good visualizations and interfaces.