Subplot does not work correctly when shareY = TRUE

[nitram9]

I wanted to create a subplot where the Y axis is shared when plotting data on the X axis. The Y axis contains strings as shown in test_df[,1].

However, since I cannot seem to force the order of the Y variable, the code (below) produces this plot:

# load library if not already done
library(plotly)

# create demo dataframe
test_df <- data.frame(c("one","two","three","four","five","six","seven","eight"), stringsAsFactors = FALSE)
colnames(test_df) <- "category"
test_df$list1 <- c( 20 , 40 , 60 , 80 , 10 , 20 , 30 , 100 )
test_df$list2 <- c( 25 , 5 , 5 , 10 , 50 , 10 , 20 , 0 )
test_df$list3 <- c( 25 , 5 , 5 , 10 , 50 , 10 , 20 , 0 )

# create subplots
p1 <- plot_ly(data= test_df, x = test_df[[2]], y= test_df[[1]], name='List 1', type='bar', marker = list(color = 'rgb(255, 128, 128)'))
p2 <- plot_ly(data= test_df,  x = test_df[[3]], y= test_df[[1]], name='List 2', type='bar', marker = list(color = 'rgb(230, 0, 0)'))
p3 <- plot_ly(data= test_df, x = test_df[[4]],  y= test_df[[1]], name='List 3', type='bar', marker = list(color = 'rgb(131, 134, 135)'))
p<-subplot(p1, p2, p3, nrows = 1, shareX = FALSE, shareY = TRUE) %>%
  layout(
      xaxis = list(title = "values 1", tickangle = 90),
      xaxis2 = list(title = "values 2", tickangle = 90),
      xaxis3 = list(title = "values 3", tickangle = 90),
      yaxis = list(title = "Categories", showgrid = TRUE),
      legend = list(orientation = 'h', y = 1.2)
    )

# generate plot
p

Since I want to order the Y axis as one, two, three, etc... and if I understand Plotly correctly, I can order the Y axis titles properly by using 1) a column from the data frame with numbers such that the order is correct, and then 2) using tickvals and ticktext to overwrite the Y axis labels.

When I try to do the first part of this, I get a plot that I think may be a bug. This is the resulting plot and the code :

# load library if not already done
library(plotly)

# create demo dataframe
test_df <- data.frame(c("one","two","three","four","five","six","seven","eight"), stringsAsFactors = FALSE)
colnames(test_df) <- "category"
test_df$list1 <- c( 20 , 40 , 60 , 80 , 10 , 20 , 30 , 100 )
test_df$list2 <- c( 25 , 5 , 5 , 10 , 50 , 10 , 20 , 0 )
test_df$list3 <- c( 25 , 5 , 5 , 10 , 50 , 10 , 20 , 0 )
# add extra list to the df here
test_df$number_list <- c(1, 2, 3, 4, 5, 6, 7, 8)

# create subplots and now use the 5th column
p1 <- plot_ly(data= test_df, x = test_df[[2]], y= test_df[[5]], name='List 1', type='bar', marker = list(color = 'rgb(255, 128, 128)'))
p2 <- plot_ly(data= test_df,  x = test_df[[3]], y= test_df[[5]], name='List 2', type='bar', marker = list(color = 'rgb(230, 0, 0)'))
p3 <- plot_ly(data= test_df, x = test_df[[4]],  y= test_df[[5]], name='List 3', type='bar', marker = list(color = 'rgb(131, 134, 135)'))
p<-subplot(p1, p2, p3, nrows = 1, shareX = FALSE, shareY = TRUE) %>%
  layout(
      xaxis = list(title = "values 1", tickangle = 90),
      xaxis2 = list(title = "values 2", tickangle = 90),
      xaxis3 = list(title = "values 3", tickangle = 90),
      yaxis = list(title = "Categories", showgrid = TRUE),
      legend = list(orientation = 'h', y = 1.2)
    )

# generate plot
p

I also tried using a single plot but the same behavior continues. Am I making an obvious mistake or is this indeed a bug?

Thanks for your help!

[trekonom]

This is not a bug, e.g. ggplot2 will give you the same plots, and not related to using subplot or a shared y-axis. In the first case you map a categorical variable on the y-axis and a numeric on the x-axis. As a sensible default you get the bars plotted horizontal by categories. In the second case both are numerics. Hence plotly (or ggplot2) don't know which var represents the categories and you get a plot of y vs. x, i.e. you get vertical bars. In this case you have to force horizontal bars by adding orientation = "h" and you get the desired result.

[nitram9]

Thank you for pointing this out @trekonom!

Perhaps on a similar issue, could you indicate whether layout(hovermode = 'compare') should or should not be working for subplots for Plotly with type = "bar"?

[trekonom]

Should work. And works. (: Check it out. Or do you any good reason to think that it isn't?

[nitram9]

@trekonom Sorry, I should have been more clear. When I use layout(hovermode = 'compare'), it doesn't give me the behavior I was expecting: to hover over the data point in each of the subplots. Instead, it only works on a single subplot.

Perhaps this is the behavior that should occur but it could be helpful to have a hover that works across the subplots when an axis is shared. Does that make sense to you?

[trekonom]

As far as I got it you have to write a custom event handler to achieve this. Following this and this add

htmlwidgets::onRender(p, readLines("multi_hover.js"))

to your script where multi_hover.js looks like this:

// Start of the multi_hover.js file
function(el) {
  el.on("plotly_hover", function(d) {
    var points = d.points[0],
        pointNum = points.pointNumber;
 
    Plotly.Fx.hover(el.id, [
    				{ curveNumber: 0, pointNumber: pointNum },
    				{ curveNumber: 1, pointNumber: pointNum },
    				{ curveNumber: 2, pointNumber: pointNum }
    				],
        			['xy', 'x2y', 'x3y']
    );
  });
}