manipulateWidget: Add Controls to Interactive Plots
In manipulateWidget: Add Even More Interactivity to Interactive Charts

Description Usage Arguments Value Advanced Usage Modify the returned widget Examples

This function permits to add controls to an interactive plot created with packages like dygraphs, highcharter or plotly in order to change the input data or the parameters of the plot.

Technically, the function starts a shiny gadget. The R session is bloqued until the user clicks on "cancel" or "done". If he clicks on "done", then the the function returns the last displayed plot so the user can modify it and/or save it.

manipulateWidget(
  .expr,
  ...,
  .updateBtn = FALSE,
  .saveBtn = TRUE,
  .exportBtn = TRUE,
  .exportType = c("html2canvas", "webshot"),
  .viewer = c("pane", "window", "browser"),
  .compare = NULL,
  .compareOpts = compareOptions(),
  .translations = mwTranslations(),
  .return = function(widget, envs) {     widget },
  .width = NULL,
  .height = NULL,
  .runApp = TRUE
)

`.expr`	expression to evaluate that returns an interactive plot of class `htmlwidget`. This expression is re-evaluated each time a control is modified.
`...`	One or more named control arguments created with functions `mwSlider`, `mwText`, etc. The name of each control is the name of the variable the controls modifies in the expression. One can also create a group of inputs by passing a list of such control arguments. for instance `mygroup = list(txt = mwText(""), nb = mwNumeric(0))` creates a group of inputs named mygroup with two inputs named "txt" and "nb".
`.updateBtn`	Should an update button be added to the controls ? If `TRUE`, then the graphic is updated only when the user clicks on the update button.
`.saveBtn`	Should an save button be added to the controls ? For saving output as html. Does not work in RStudio Viewer
`.exportBtn`	Should an export button be added to the controls ? For saving output as png. Does not work in RStudio Viewer
`.exportType`	`.exportBtn`, using `html2canvas` (default) and keeping current zoom, ... or using `webshot`
`.viewer`	Controls where the gadget should be displayed. `"pane"` corresponds to the Rstudio viewer, `"window"` to a dialog window, and `"browser"` to an external web browser.
`.compare`	Sometimes one wants to compare the same chart but with two different sets of parameters. This is the purpose of this argument. It can be a character vector of input names or a named list whose names are the names of the inputs that should vary between the two charts. Each element of the list must be a vector or a list of length equal to the number of charts with the initial values of the corresponding parameter for each chart. It can also be `NULL`. In this case, the parameter is initialized with the default value for the two charts.
`.compareOpts`	List of options created `compareOptions`. These options indicate the number of charts to create and their disposition.
`.translations`	List of translation strings created with function `mwTranslations`. Used to translate UI titles and labels.
`.return`	A function that can be used to modify the output of `manipulateWidget`. It must take two parameters: the first one is the final widget, the second one is a list of environments containing the input values of each individual widget. The length of this list is one if .compare is null, two or more if it has been defined.
`.width`	Width of the UI. Used only on Rmarkdown documents with option `runtime: shiny`.
`.height`	Height of the UI. Used only on Rmarkdown documents with option `runtime: shiny`.
`.runApp`	(advanced usage) If true, a shiny gadget is started. If false, the function returns a `MWController` object. This object can be used to check with command line instructions the behavior of the application. (See help page of `MWController`). Notice that this parameter is always false in a non-interactive session (for instance when running tests of a package).

The result of the expression evaluated with the last values of the controls. It should be an object of class htmlWidget.

The "normal" use of the function is to provide an expression that always return an htmlwidget. In such case, every time the user changes the value of an input, the current widget is destroyed and a new one is created and rendered.

Some packages provide functions to update a widget that has already been rendered. This is the case for instance for package leaflet with the function leafletProxy. To use such functions, manipulateWidget evaluates the parameter .expr with four extra variables:

.initial: TRUE if the expression is evaluated for the first time and then the widget has not been rendered yet, FALSE if the widget has already been rendered.
.session: A shiny session object.
.output: ID of the output in the shiny interface.
.id: Id of the chart. It can be used in comparison mode to make further customization without the need to create additional input controls.

You can take a look at the last example to see how to use these two variables to update a leaflet widget.

In some specific situations, a developer may want to use manipulateWidget in a function that waits the user to click on the "Done" button and modifies the widget returned by manipulateWidget. In such situation, parameter .return should be used so that manipulateWidget is the last function called. Indeed, if other code is present after, the custom function will act very weird in a Rmarkdown document with "runtime: shiny".

# Basic example with fake data
if (require(dygraphs)) {
  mydata <- data.frame(period = 1:100, value = rnorm(100))
  manipulateWidget(dygraph(mydata[range[1]:range[2], ], main = title),
                   range = mwSlider(1, 100, c(1, 100)),
                   title = mwText("Fictive time series"))
}

# Let use manipulateWidget to explore the evolution of energy consumption in
# the world
data("worldEnergyUse")

if (require(plotly)) {
  # Function that generates a chart representing the evolution of energy
  # consumption per country. Creating a function is not necessary. We do it
  # for clarity and reuse in the different examples.
  plotEnergyUse <- function(Country, Period, lwd = 2, col = "gray") {
    dataset <- subset(
      worldEnergyUse,
      country == Country & year >= Period[1] & year <= Period[2]
    )
    plot_ly(dataset) %>%
      add_lines(~year, ~energy_used, line = list(width = lwd, color = col)) %>%
      layout(title = paste("Energy used in", Country))
  }

  # Launch the interactive visualisation
  manipulateWidget(
    plotEnergyUse(Country, Period),
    Period = mwSlider(1960, 2014, c(1960, 2014)),
    Country = mwSelect(sort(unique(worldEnergyUse$country)), "United States")
  )

  # Directly start comparison mode
  manipulateWidget(
    plotEnergyUse(Country, Period),
    Period = mwSlider(1960, 2014, c(1960, 2014)),
    Country = mwSelect(sort(unique(worldEnergyUse$country))),
    .compare = list(Country = c("United States", "China")),
    .compareOpts = compareOptions(ncol = 2)
  )

  # Dynamic input parameters
  #-------------------------
  # The arguments of an input can depend on the values of other inputs.
  # In this example, when the user changes the region, the choices of input
  # "Country" are updated with the countries of that region.

  # First we create a list that contains for each region the countries in that
  # retion
  refRegions <- by(worldEnergyUse$country, worldEnergyUse$region,
                   function(x) as.character(sort(unique(x))))

  manipulateWidget(
    plotEnergyUse(Country, Period),
    Period = mwSlider(1960, 2014, c(1960, 2014)),
    Region = mwSelect(sort(unique(worldEnergyUse$region))),
    Country = mwSelect(choices = refRegions[[Region]])
  )

  # Grouping inputs
  #----------------
  # Inputs can be visually grouped with function mwGroup()
  manipulateWidget(
    plotEnergyUse(Country, Period, lwd, col),
    Period = mwSlider(1960, 2014, c(1960, 2014)),
    Country = mwSelect(sort(unique(worldEnergyUse$country)), "United States"),
    `Graphical Parameters` = mwGroup(
      lwd = mwSlider(1,10, 2, label = "Line Width"),
      col = mwSelect(choices = c("gray", "black", "red"))
    )
  )

  # Conditional inputs
  #-------------------
  # Inputs can be displayed or hidden depending on the state of other inputs.
  # In this example, user can choose to display the level of aggregation
  # (region or country). Depending on the choixe, the application displays
  # input Region or input Country.
  plotEnergyUseRegion <- function(Region, Period, lwd = 2, col = "gray") {
    dataset <- subset(
      worldEnergyUse,
      region == Region & year >= Period[1] & year <= Period[2]
    )
    dataset <- aggregate(energy_used ~ year, sum, data = dataset)

    plot_ly(dataset) %>%
      add_lines(~year, ~energy_used, line = list(width = lwd, color = col)) %>%
      layout(title = paste("Energy used in", Region))
  }

  manipulateWidget(
    {
      if (Level == "Region") {
        plotEnergyUseRegion(Region, Period)
      } else {
        plotEnergyUse(Country, Period)
      }
    },
    Period = mwSlider(1960, 2014, c(1960, 2014)),
    Level = mwSelect(c("Region", "Country")),
    Region = mwSelect(sort(unique(worldEnergyUse$region)),
                      .display = Level == "Region"),
    Country = mwSelect(sort(unique(worldEnergyUse$country)),
                       .display = Level == "Country")
  )

}

# Advanced Usage
# --------------
# When .expr is evaluated with tehnical variables:
# .initial: is it the first evaluation?
# .outputId: integer representing the id of the chart
# .output: shiny output id
# .session: shiny session
# They can be used to update an already rendered widget instead of replacing
# it each time an input value is modified.
#
# In this example, we represent on a map, the energy use of countries.
# When the user changes an input, the map is not redrawn. Only the circle
# markers are updated.
if (require(leaflet)) {
  plotMap <- function(Year, MaxRadius = 30, .initial, .session, .output) {
    dataset <- subset(worldEnergyUse, year == Year)
    radius <- sqrt(dataset$energy_used) /
      max(sqrt(worldEnergyUse$energy_used), na.rm = TRUE) * MaxRadius

    if (.initial) { # map has not been rendered yet
      map <- leaflet() %>% addTiles()
    } else { # map already rendered
      map <- leafletProxy(.output, .session) %>% clearMarkers()
    }

    map %>% addCircleMarkers(dataset$long, dataset$lat, radius = radius,
                             color = "gray", weight = 0, fillOpacity = 0.7)
  }

  manipulateWidget(
    plotMap(Year, MaxRadius, .initial, .session, .output),
    Year = mwSlider(1960, 2014, 2014),
    MaxRadius = mwSlider(10, 50, 20)
  )
}