Create download handler for plot with source data

Generates a Shiny downloadHandler() that bundles the interactive plot, images of it, and its supporting data into a single .zip archive.

Usage

create_source_download_handler(
  data_list,
  filename_base = "source_data",
  images = TRUE,
  output_id = "download.source",
  session = getDefaultReactiveDomain()
)

Arguments

data_list: A reactive returning either a single summary list produced by collect_source_data() (with elements plot, plot_data, stats, and inputs), or a named list of such summaries (one per plot). When a named list of summaries is supplied, each summary is written to its own set of files (prefixed with the list name) so several plots can be bundled into a single archive.
filename_base: character(1). Base name for the downloaded .zip file without extension. The final filename takes the form <filename_base>_<Sys.Date()>.zip.
images: logical(1). Whether to include .svg and .png images of each plot. Requires the button to carry the markup module_tack_ui() gives it; a hand-rolled shiny::downloadButton() without it simply gets no images.
output_id: character(1). The id this handler is assigned to within its module, used to find the images the browser sends. Only needs changing if the handler is assigned to something other than output$download.source.
session: The module session. Defaults to the calling module's, which is what every in-package call site wants.

Value

A downloadHandler object suitable for assignment to a Shiny output.

Details

The archive holds, per plot: the interactive plot as self-contained HTML (<name>_plot.html), an <name>_plot.svg and <name>_plot.png of it, and CSVs of the plot data, the statistics, and the UI inputs.

The images are photographed in the browser, off the graph the user is looking at, so they carry every edit made after the figure was built – reference lines, statistical brackets, restyled axes and legends, dragged annotations. The capture happens between the button's click and the download itself, which is why the button pauses briefly before the archive arrives. When it cannot be done – the capture fails, the round trip times out, the plot sits on a hidden tab – the archive still downloads, without the images.

A module whose output is not a plotly graph has nothing for the browser to photograph and draws itself instead, by putting a vector_svg and/or raster_png function of (width, height, res) on its summary list, or on the reactive passed as data_list. draw_to_svg() and draw_to_png() build one from any grid or base drawing; ComplexHeatmap_HeatmapServer() is the worked example.

A summary may also carry svg_key, naming the captured image that belongs to it. The Figure Builder uses this because the browser knows a panel only by its id, while the summary is named after the panel's display label.

One caveat worth passing on to users: a plot drawn with WebGL (the dittoViz scatter plot's WebGL toggle) can only be photographed as a raster, so its points arrive as an embedded image inside an otherwise vector SVG. Turning WebGL off gives a fully editable file.

Author

Jacob Martin, Jared Andrews

Examples

if (FALSE) { # \dontrun{
# Example usage in a Shiny app
library(shiny)
library(plotly)
library(VizModules)
ui <- fluidPage(
    plotlyOutput("my_plot"),
    downloadButton("download_data", "Download Plot and Data")
)

server <- function(input, output) {
    plot_reactive <- reactive({
        plot_ly(mtcars, x = ~mpg, y = ~hp, type = "scatter", mode = "markers")
    })

    output$my_plot <- renderPlotly(plot_reactive())
    # collect_source_data() reads reactives, so it has to run inside one.
    inputs_reactive <- reactive(reactiveValuesToList(input))
    output$download_data <- create_source_download_handler(
        reactive(collect_source_data(plot_reactive, inputs_reactive = inputs_reactive))
    )
}

shinyApp(ui, server)
} # }