In my last article on the STATWORX blog, I have guided you through the process of writing a complex JavaScript-callback. However, most users might be slightly frustrated by the lack of arguments to customize a standard rBokeh plot fully. Actually, rBokeh is a little bit outdated (structure() warnings all the way!) and lacks some functionalities that are available in its Python equivalent. But don’t toss in the towel right away! I created some workarounds for these, which you hopefully find helpful.

In general

This approach is my go-to solution to change a rBokeh plot for which there is an argument missing in rBokeh that is available in python.

• Create the plot.
• Inspect the structure (str(plot)) of the rBokeh object.
• Search for the python’s argument name.
• Overwrite the value with the desired option as derived from python’s bokeh.

So, first of all, I set up an initial rBokeh figure that we manipulate later on.

plot <- figure(data = iris) %>% 
ly_bar(x = Species,
y = Sepal.Length,
hover = TRUE)

Manipulate the hover functionality

The first set of tricks deals with the customization of hover effects. Hover effects are essentials of interactive plots, so it makes a lot of sense to invest some time in optimizing them.

Anchor

Unlike in python’s bokeh, there is no anchor argument to change the position of a hover tooltip. By default, it appears in the center of the hovered element. To change it, we need to deep dive into the rBokeh object. The object is a deep and complex nested list in which all the information about the plot is stored. While some elements are always structured in the same way, different layers are named by a seemingly arbitrary string (e.g., 51dab389c6209bbf084a86b368f68724). I wrote the following code snippet to change the hover position from center to top_center.

# Get the position of the anchor argument within the object-list
xyz <- logical()
for (i in seq_along(plot$x$spec$model)) {
  xyz[i] <- !is.null(plot$x$spec$model[[i]]$attributes$point_policy)
}

# Solution using for loop
for (i in which(xyz)) {
plot$x$spec$model[[i]]$attributes$anchor <- "top_center"
}

In case you are not very fond of simple for loops, here are also solutions with purrr or lapply:

# Solution using purrr
xyz <- purrr::map_lgl(plot$x$spec$model, .f = ~ !is.null(.x$attributes$anchor))

plot$x$spec$model[which(xyz)] <- purrr::map(plot$x$spec$model[which(xyz)], 
~{
.$attributes$anchor <- "top_center"
return(.)
})

# Solution using the apply family
xzy <- sapply(plot$x$spec$model, function(x) !is.null(x$attributes$anchor))
plot$x$spec$model[which(xyz)] <- lapply(plot$x$spec$model[xyz], 
function(abc) {
abc$attributes$anchor <- "top_center"
return(abc)
})

All options of the tooltip position can be found here.

Point policy

Another option that can be specified in the same way is whether the tooltip should appear at a specific place (snap_to_data) or should follow the courser (follow_mouse). This point_policy option is also missing in rBokeh but can be added by the same logic. Here is a solution for the purrr way but all other descriped options work as well.

# Get the position of the point policy argument within the object-list
xyz <- purrr::map_lgl(plot$x$spec$model, .f = ~ !is.null(.x$attributes$point_policy))

plot$x$spec$model[which(xyz)] <- purrr::map(plot$x$spec$model[which(xyz)], 
~{
.$attributes$point_policy <- "follow_mouse"
return(.)
})

What you see is what you want

The last hover-related issue I want to address are the shown values. rBokeh is rather inflexible in this context. Sometimes (e.g., in ly_points) it is possible to define a specific hover information (either a variable from the data or another data frame/list of the same length as the plot data) but in other cases the hover argument is just logical (TRUE or FALSE, like in ly_bar). If you want to change its default tooltip you need to do this by hand, again.

# Set up the figure
plot <- figure(data = iris) %>% 
ly_bar(x = Species,
y = Sepal.Length,
hover = T)

# get the list elements where tooltips are defined
hover_info <- purrr::map_lgl(plot$x$spec$model, .f = ~ !is.null(.x$attributes$tooltips))

# delete a specific tooltip
plot$x$spec$model[[which(hover_info)]]$attributes$tooltips[[2]] <- NULL

# add a tooltip
plot$x$spec$model[[which(hover_info)]]$attributes$tooltips[[2]] <- 
# list of printed name (test) and name for internal use (@hover_col_3)
list("test","@hover_col_3")

hover_data <- purrr::map_lgl(plot$x$spec$model, .f = ~ !is.null(.x$attributes$data$hover_col_1))

# manipulate a tooltip
plot$x$spec$model[which(hover_data)] <- purrr::map(plot$x$spec$model[which(hover_data)], 
~{
.$attributes$data$hover_col_1 <- 1:3
# must match assigned name above
.$attributes$data$hover_col_3 <- letters[1:3]
return(.)
})

plot

Keep plotting!

I hope you enjoyed my blog post, and it helps you in solving or avoiding some troubles with rBokeh. And who knows, maybe a more intense use of this package might even motivate the developers to update or further develop this excellent package. So, keep plotting!

All options of the tooltip position can be found here.

In case you are not very fond of simple for loops, here are also solutions with purrr or lapply.

Über den Autor

Matthias Nistler

I am a data scientist at STATWORX and passionate for wrangling data and getting the most out of it. Outside of the office, I use every second for cycling until the sun goes down.

Der Beitrag rBokeh – Don't be stopped by missing arguments! erschien zuerst auf STATWORX.