Upcoming ggplot2 version incompatible with BCEA

[teunbrand]

Hi there,

We've been preparing a new major release for ggplot2 and found an issue during a reverse dependency check.
The issue is a broken example and it occurred in the following place:

BCEA/R/plot.bcea.R

Lines 76 to 81 in 152c2b4

	#' plot(he,
	#' graph = "ggplot2", # use ggplot2
	#' ICER_size = 1.5, # extra options modifies the mean point size
	#' text = list(size=rel(1.25)) # modifies the text size
	#' )
	#' }

More specifically, I think the text element becomes corrupted around this step in the plotting function.

BCEA/R/plot.bcea.R

Lines 166 to 169 in 152c2b4

	global_params <-
	modifyList(default_params,
	extra_params,
	keep.null = TRUE)

You can test your code with the development version of ggplot2 by installing it as follows:

# install.packages("pak")
pak::pak("tidyverse/ggplot2")

We aim to release the new ggplot2 version in about 2 weeks, and hope you can submit a fix to CRAN around that time. Hopefully this will inform you in a timely manner.

Best wishes,
Teun

[giabaio]

Thanks, @teunbrand. We've been working on some relatively major changes, which I have just integrated into the main repo. I'll look into this and hopefully be able to fix quickly.

[giabaio]

OK, so. I can see the error:

plot(he,graph = "gg",ICER_size=5.5,text=list(size=rel(1.25)))
Error in `method(update_ggplot, list(ggplot2::theme, ggplot2::ggplot))`:
! Can't merge the `text` theme element.
Caused by error in `method(merge_element, list(class_any, class_any))`:
! No method for merging <list> into <ggplot2::element_text>.
Run `rlang::last_trace()` to see where the error occurred.

I think the problem is that we define a set of default parameters

default_params <- 
        list(text = element_text(size = 9),
             legend.key.size = grid::unit(0.5, "lines"),
             legend.spacing = grid::unit(-1.25, "line"),
             panel.grid = element_blank(),
             legend.key = element_blank(),
             plot.title = element_text(
               lineheight = 1,
               face = "bold",
               size = 11.5,
               hjust = 0.5))

and then allow for some of them to be rewritten as

keep_param <-
        names(default_params)[names(default_params) %in% names(extra_args)]
      
      extra_params <- extra_args[keep_param]
      
      global_params <-
        modifyList(default_params,
                   extra_params,
                   keep.null = TRUE)
      
      theme_add <- Filter(f = \(val) ggplot2::is_theme(val), x = extra_args)
      
      ceplane <-
        ceplane.plot(x,
                     wtp = wtp,
                     pos = pos,
                     comparison = comparison,
                     graph = "ggplot2", ...) +
        do.call(theme, global_params) +
        theme_add

I think the issue is in how the new S7 object handles the addition of a theme? @n8thangreen is this something obvious to you? @teunbrand is this the problem, do you think?

[teunbrand]

IIRC there was a text element somewhere that was just <element_name> = list(size = x) (which is corrupted) instead of <element_name> = element_text(size = x) that was created in the construction of the global_params variable.

I'm not sure if this also happened with the previous version of ggplot2, but if it did, it only worked by mistake.

To explain a little bit more:
We cannot add two theme objects if one (or both) of them contains corrupted elements. The reason for this is that in adding two themes with +, you're transferring properties from the old theme to the new theme. It fills in NULL properties from the new element with populated properties from the old element. If one is corrupted, it is impossible to determine what properties should be transferred.

[teunbrand]

If it is hard to figure out, I can try to tinker on my own and sent a PR your way

[giabaio]

Thanks. Let us have a go --- I had a very quick look earlier and it looked like the theme was kind of set. In the old ggplot, the list would modify the size of the text element, indeed. With the new one, it breaks the code, but also

plot(he,graph = "gg") + theme(text=element_text(size=...))

doesn't seem to have any effect (which I thought would be an easy fix?...).

It may be that I've not tried properly and will look more closely as soon as I get a minute!

[teunbrand]

If you find any additional issues, such as adding themes not working in ggplot2 itself, please keep us in the loop and we can see if something should be done in ggplot2 itself.

[giabaio]

So: the offending code acts to create a multiplot made by 4 different types of graphs, each called by a BCEA function. For instance, the first of the 4 panels is produced by calling the function ceplane.plot. Interestingly, if I do

ceplane.plot(he,graph="gg",ICER_size=6,text=list(size=rel(3)))

that works with no issues... So is it the problem to do with multiplot?

[n8thangreen]

Sorry but I don't have any good ideas about this other than the element_text thing!

[giabaio]

This seems to be doing it... (give or take... I've only modified the ceplane.plot for now and the legend is not quite right -- but I suspect I'm missing some key formatting of what's now theme_BCEA). Does this sound like a decent idea? (@n8thangreen)

#' A Custom ggplot2 Theme for BCEA
#'
#' This theme is designed for use with plots created using the `BCEA` package. 
#' It builds on `theme_bw()` and customizes axis text, titles, 
#' plot background, and legend styling.
#'
#' @return A ggplot2 theme object that can be added to a ggplot.
#' @examples
#' library(ggplot2)
#' library(BCEA)
#' ggplot(mtcars, aes(wt, mpg)) +
#'  geom_point() +
#'  theme_BCEA() 
#' @export
#' 
theme_BCEA <- function() {
  theme_bw() +
    theme(
      text = element_text(size = 9),
      legend.key.size = grid::unit(0.5, "lines"),
      legend.spacing = grid::unit(-1.25, "line"),
      panel.grid = element_blank(),
      legend.key = element_blank(),
      plot.title = element_text(
        lineheight = 1,
        face = "bold",
        size = 11.5,
        hjust = 0.5
      )
    )
}

#' Utility function to modify simply the `ggplot` `theme_BCEA()` theme
#' using inputs inside the call to `plot.bcea`
#'
#' @noRd
#' @keywords internal
#' 
process_element_arg <- function(arg, constructor) {
  
  # Define a safe resolve_rel_size() helper
  resolve_rel_size <- function(size, base = 9) {
    if (inherits(size, "rel") && !is.null(attr(size, "val"))) {
      return(attr(size, "val") * base)
    } else if (is.numeric(size)) {
      return(size)
    } else {
      stop("Invalid size value: must be numeric or rel().")
    }
  }
  
  if (inherits(arg, class(constructor()))) {
    return(arg)
  } else if (is.list(arg)) {
    if ("size" %in% names(arg)) {
      arg$size <- resolve_rel_size(arg$size, base = base_size)
    }
    return(do.call(constructor, arg))
  } else if (is.null(arg)) {
    return(NULL)
  } else {
    stop("Invalid theme element.")
  }
}


#' Summary Plot of the Health Economic Analysis
#' 
#' Plots in a single graph the Cost-Effectiveness plane, the Expected
#' Incremental Benefit, the CEAC and the EVPI.
#' 
#' The default position of the legend for the cost-effectiveness plane
#' (produced by [ceplane.plot()]) is set to `c(1, 1.025)`
#' overriding its default for `pos=FALSE`, since multiple ggplot2 plots
#' are rendered in a slightly different way than single plots.
#' 
#' @param x A `bcea` object containing the results of the Bayesian
#' modelling and the economic evaluation.
#' @template args-comparison
#' @param wtp The value of the willingness to pay parameter. It is passed to
#' [ceplane.plot()].
#' @template args-pos
#' @param graph A string used to select the graphical engine to use for
#' plotting. Should (partial-)match the options `"base"`,
#' `"ggplot2"` or `"plotly"`. Default value is `"base"`.
#' @param ...  Arguments to be passed to the methods [ceplane.plot()]
#' and [eib.plot()]. Please see the manual pages for the individual
#' functions.  Arguments like `size`, `ICER.size` and `plot.cri`
#' can be supplied to the functions in this way. In addition if
#' `graph="ggplot2"` and the arguments are named theme objects they will
#' be added to each plot.
#' 
#' @return A plot with four graphical summaries of the health economic evaluation.
#' 
#' @author Gianluca Baio, Andrea Berardi
#' 
#' @seealso [bcea()],
#'          [ceplane.plot()],
#'          [eib.plot()],
#'          [ceac.plot()],
#'          [evi.plot()]
#' @importFrom Rdpack reprompt
#'          
#' @references
#' 
#' \insertRef{Baio2011}{BCEA}
#' 
#' \insertRef{Baio2013}{BCEA}
#' 
#' @keywords hplot
#' 
#' @examples
#' # See Baio G., Dawid A.P. (2011) for a detailed description of the 
#' # Bayesian model and economic problem
#'
#' # Load the processed results of the MCMC simulation model
#' data(Vaccine)
#' 
#' # Runs the health economic evaluation using BCEA
#' he <- bcea(
#'        e=eff, c=cost,        # defines the variables of 
#'                              #  effectiveness and cost
#'        ref=2,                # selects the 2nd row of (e,c) 
#'                              #  as containing the reference intervention
#'        interventions=treats, # defines the labels to be associated 
#'                              #  with each intervention
#'        Kmax=50000,           # maximum value possible for the willingness 
#'                              #  to pay threshold; implies that k is chosen 
#'                              #  in a grid from the interval (0,Kmax)
#'        plot=FALSE            # does not produce graphical outputs
#'       )
#'
#' # Plots the summary plots for the "bcea" object m using base graphics
#' plot(he, graph = "base")
#' 
#' # Plots the same summary plots using ggplot2
#' if(require(ggplot2)){
#' plot(he, graph = "ggplot2")
#' 
#' # Example of a customized plot.bcea with ggplot2
#' plot(he,
#'   graph = "ggplot2",          # use ggplot2
#'   ICER_size = 1.5,            # extra options modifies the mean point size
#'   text = list(size=rel(1.25)) # modifies the text size
#'   )                                                       
#' }
#' 
#' @import ggplot2
#' @export
#' 
plot.bcea <- function(x,
                      comparison = NULL,
                      wtp = 25000,
                      pos = FALSE,
                      graph = c("base", "ggplot2", "plotly"),
                      ...) {
  
  
  graph <- match.arg(graph)
  extra_args <- list(...)
  
  # consistent colours across plots
  if (is.element("point", names(extra_args))) {
    if (is.element("color", names(extra_args$point))) {
      extra_args$line$color <- extra_args$point$color   
    }}
  
  if (is_baseplot(graph)) {
    withr::with_par(list(mfrow = c(2,2)), {
      ceplane.plot(x,
                   comparison = comparison,
                   wtp = wtp,
                   pos = pos,
                   graph = "base",...)
      
      do.call(eib.plot,
              c(he = list(x),
                pos = pos,
                comparison = comparison,
                graph = "base",
                extra_args))
      
      do.call(ceac.plot,
              c(he = list(x),
                pos = pos,
                graph = "base",
                extra_args))
      
      evi.plot(x,
               graph = "base", ...)
    })
  } else if (is_ggplot(graph)) {
    
    is_req_pkgs <- unname(sapply(c("ggplot2", "grid"),
                                 requireNamespace, quietly = TRUE))
    
    if (!all(is_req_pkgs)) {
      message("falling back to base graphics\n")
      plot.bcea(
        x,
        comparison = comparison,
        wtp = wtp,
        pos = pos,
        graph = "base", ...)
      
      return(invisible(NULL))
    }
    
    if (all(is_req_pkgs)) {
      
      # Map theme element names to their constructors
      element_constructors <- list(
        text = ggplot2::element_text,
        axis.text = ggplot2::element_text,
        axis.title = ggplot2::element_text,
        legend.title = ggplot2::element_text,
        legend.text = ggplot2::element_text,
        plot.title = ggplot2::element_text,
        strip.text = ggplot2::element_text,
        line = ggplot2::element_line,
        axis.line = ggplot2::element_line,
        rect = ggplot2::element_rect,
        panel.background = ggplot2::element_rect
        # Add more as needed
      )
      
      # Process only matching theme elements
      for (nm in names(extra_args)) {
        if (nm %in% names(element_constructors)) {
          constructor <- element_constructors[[nm]]
          update_theme_args[[nm]] <- process_element_arg(
            extra_args[[nm]], constructor
          )
        }
      }
      
      ceplane <-
        ceplane.plot(x,
                     wtp = wtp,
                     pos = pos,
                     comparison = comparison,
                     graph = "ggplot2", ...) + theme_BCEA() +
        do.call(ggplot2::theme, update_theme_args)
      
      eib <-
        do.call(eib.plot,
                c(he = list(x),
                  pos = pos,
                  comparison = comparison,
                  graph = "ggplot2",
                  extra_args))
      # +
      #   do.call(theme, global_params) +
      #   theme_add
      
      ceac <-
        do.call(ceac.plot,
                c(he = list(x),
                  pos = pos,
                  comparison = comparison,
                  graph = "ggplot2",
                  extra_args)) 
      # +
      #   do.call(theme, global_params) +
      #   theme_add
      
      evi <-
        evi.plot(x, graph = "ggplot2", ...) 
      # +
      #   do.call(theme, global_params) +
      #   theme_add
      
      multiplot(list(ceplane, ceac, eib, evi),
                cols = 2)
    }
  } else if (is_plotly(graph)) {
    extract_plotly_title = function(p) {
      if (exists("layoutAttrs", p$x)) {
        return(p$x$layoutAttrs[[1]]$title)
      } else if (exists("layout", p$x)) {
        return(p$x$layout$title)
      } else {
        message("Plotly title not found in extract_plotly_title()")
        return("Plotly title not found")
      }
    }
    
    p1 = ceplane.plot(x, graph = "p", wtp = wtp, ...)
    p2 = eib.plot(x, graph = "p", ...)
    p3 = ceac.plot(x, graph = "p", ...)
    p4 = evi.plot(x, graph = "p", ...)
    
    ppp = plotly::subplot(
      p1, p2, p3, p4, nrows = 2,
      titleX = TRUE, titleY = TRUE, margin = 0.1
    )
    
    annotations = list(
      list(
        x = 0.2, y = 1.0,
        text = extract_plotly_title(p1),
        xref = "paper",
        yref = "paper",
        xanchor = "center",
        yanchor = "bottom",
        showarrow = FALSE
      ),
      list(
        x = 0.8, y = 1.0,
        text = extract_plotly_title(p2),
        xref = "paper",
        yref = "paper",
        xanchor = "center",
        yanchor = "bottom",
        showarrow = FALSE
      ),
      list(
        x = 0.2, y = 0.4,
        text = extract_plotly_title(p3),
        xref = "paper",
        yref = "paper",
        xanchor = "center",
        yanchor = "bottom",
        showarrow = FALSE
      ),
      list(
        x = 0.8, y = 0.4,
        text = extract_plotly_title(p4),
        xref = "paper",
        yref = "paper",
        xanchor = "center",
        yanchor = "bottom",
        showarrow = FALSE
      )
    )
    
    ppp = ppp |>
      layout(annotations = annotations, title = "")
    return(ppp)
  }
}

[giabaio]

@teunbrand we have something that is at least OK to pass the tests and behave with the new version of ggplot2. We're hoping to find some time to make that code a bit nicer (as opposed to just working) -- bear with us (both @n8thangreen and I are teaching in a summer school this week, so may take a little to do this). But it'll be done.

[teunbrand]

Thanks Gianluca for the update! We're still at least 2 weeks away from submitting to CRAN, so there is a bit of time.

[giabaio]

[like] Baio, Gianluca reacted to your message:

…

________________________________ From: Teun van den Brand ***@***.***> Sent: Monday, June 23, 2025 11:16:03 AM To: giabaio/BCEA ***@***.***> Cc: Baio, Gianluca ***@***.***>; Comment ***@***.***> Subject: Re: [giabaio/BCEA] Upcoming ggplot2 version incompatible with BCEA (Issue #74) ⚠ Caution: External sender [https://avatars.githubusercontent.com/u/49372158?s=20&v=4]teunbrand left a comment (giabaio/BCEA#74)<#74 (comment)> Thanks Gianluca for the update! We're still at least 2 weeks away from submitting to CRAN, so there is a bit of time. — Reply to this email directly, view it on GitHub<#74 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACCRVSM2VVTAHABCHEU7FHD3E7OXHAVCNFSM6AAAAAB7NKRBG6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSOJWGAYTENZWHA>. You are receiving this because you commented.Message ID: ***@***.***>