ggplot2 v.3.3.0: Error in (function (el, elname) : "plot.title.position" is not a valid theme element name.

[thierrygosselin]

I'm having an error in my code that used to work prior to upgrading to ggplot2 v.3.3.0. It says:

Error in (function (el, elname)  :    "plot.title.position" is not a valid theme element name.

The problem is that plot.title.position is new to v.3.3.0 it's not used in my code to produce the box plot. When I'm trying to do the function line by line to reproduce the problem and find exactly where the code breaks it doesn't give me an error. Very puzzling. So a short reprex at this time is not possible. I'm working on it.

However, is this something seen before ?

Thierry

[clauswilke]

Are you somehow mixing code from different ggplot2 versions? E.g., did you save a plot as an .Rda file with a previous version of ggplot2 and are now trying to render it with an updated ggplot2?

[thierrygosselin]

No, rebuilding the package with the new ggplot2 v.3.3.0, running a code that worked before, now breaks with the error above. Here is the traceback:

 Error in (function (el, elname)  : 
  "plot.title.position" is not a valid theme element name. 
11. stop("\"", elname, "\" is not a valid theme element name.") 
10. (function (el, elname) 
{
    eldef <- ggint$.element_tree[[elname]]
    if (is.null(eldef)) { ... 
9. mapply(validate_element, elements, names(elements)) 
8. (function (line, rect, text, title, aspect.ratio, axis.title, 
    axis.title.x, axis.title.x.top, axis.title.y, axis.title.y.right, 
    axis.text, axis.text.x, axis.text.x.top, axis.text.y, axis.text.y.right, 
    axis.ticks, axis.ticks.x, axis.ticks.y, axis.ticks.length,  ... 
7. do.call(theme, c(oldtheme, complete = isTRUE(attr(oldtheme, "complete")), 
    validate = old.validate & new.validate)) 
6. update_theme(p$theme, object) 
5. add_ggplot(e1, e2, e2name) 
4. `+.gg`(fig.boxplot, ggplot2::theme(axis.title.x = ggplot2::element_blank(), 
    axis.text.x = ggplot2::element_blank(), axis.ticks.x = ggplot2::element_blank())) at   visualization.R#319
3. boxplot_stats(data = rep.stats, title = "DArT marker's reproducibility", 
    subtitle = stringi::stri_join("\nLower outlier: ", outlier.rep), 
    x.axis.title = NULL, y.axis.title = "Reproducibility (proportion)", 
    bp.filename = bp.filename, path.folder = path.folder) at filter_dart_reproducibility.R#191
2.filter_dart_reproducibility(data = gds, interactive.filter = interactive.filter, 
    filter.reproducibility = filter.reproducibility, parallel.core = parallel.core, 
    verbose = verbose, parameters = filters.parameters, path.folder = wf, 
    internal = FALSE) at filter_rad.R#639
1.radiator::filter_rad(data = "BET_SEQ_COUNTS_DTu19-4053.csv", 
    strata = "bet_strata_dart.tsv") 

[thierrygosselin]

The line 319 in visualization.R#319 above is here:

fig.boxplot <- fig.boxplot +
      ggplot2::theme(
        axis.title.x = ggplot2::element_blank(),
        axis.text.x = ggplot2::element_blank(),
        axis.ticks.x = ggplot2::element_blank()
      )

This is the entire function used :

# boxplot of stats
#' @title boxplot_stats
#' @description Generate a boxplot
#' @rdname boxplot_stats
#' @keywords internal
#' @export
boxplot_stats <- function(
  data,
  title,
  subtitle = NULL,
  x.axis.title = NULL,
  y.axis.title,
  facet.columns = FALSE,
  facet.rows = FALSE,
  bp.filename = NULL,
  path.folder = NULL
) {
   if (is.null(path.folder)) path.folder <- getwd()

  n.group <- dplyr::n_distinct(data$GROUP)
  element.text <- ggplot2::element_text(size = 10,
                                        face = "bold")

  if (facet.columns) {
    data <- dplyr::mutate(data, X = "1")
    fig.boxplot <- ggplot2::ggplot(data = data, ggplot2::aes(X))
  } else {
    fig.boxplot <- ggplot2::ggplot(data = data, ggplot2::aes(GROUP))
  }


  fig.boxplot <- fig.boxplot +
    ggplot2::geom_boxplot(
      ggplot2::aes(ymin = OUTLIERS_LOW, lower = Q25, middle = MEDIAN, upper = Q75,
                   ymax = OUTLIERS_HIGH), stat = "identity") +
    ggplot2::labs(y = y.axis.title, title = title)

  if (!is.null(subtitle)) fig.boxplot <- fig.boxplot + ggplot2::labs(subtitle = subtitle)

  # Draw upper outliers
  if (facet.columns) {
    fig.boxplot <- fig.boxplot +
      ggplot2::geom_segment(
        ggplot2::aes(
          x = "1", xend = "1",
          y = OUTLIERS_HIGH, yend = MAX),
        linetype = "dashed")
  } else {
    fig.boxplot <- fig.boxplot +
      ggplot2::geom_segment(
        ggplot2::aes(
          x = GROUP, xend = GROUP,
          y = OUTLIERS_HIGH, yend = MAX),
        linetype = "dashed")
  }

  # Draw lower outliers
  if (facet.columns) {
    fig.boxplot <- fig.boxplot +
      ggplot2::geom_segment(
        ggplot2::aes(
          x = "1", xend = "1",
          y = OUTLIERS_LOW, yend = MIN),
        linetype = "dashed")
  } else {
    fig.boxplot <- fig.boxplot +
      ggplot2::geom_segment(
        ggplot2::aes(
          x = GROUP, xend = GROUP,
          y = OUTLIERS_LOW, yend = MIN),
        linetype = "dashed")
  }

  fig.boxplot <- fig.boxplot +
    ggplot2::theme(
      plot.title = ggplot2::element_text(size = 12, face = "bold", hjust = 0.5),
      legend.position = "none",
      axis.title.y = element.text,
      axis.text.y = element.text
    ) +
    ggplot2::theme_bw()

  if (is.null(x.axis.title)) {
    fig.boxplot <- fig.boxplot +
      ggplot2::theme(
        axis.title.x = ggplot2::element_blank(),
        axis.text.x = ggplot2::element_blank(),
        axis.ticks.x = ggplot2::element_blank()
      )
  } else {
    fig.boxplot <- fig.boxplot +
      ggplot2::xlab(x.axis.title) +
      ggplot2::theme(
        axis.title.x = element.text,
        axis.ticks.x = ggplot2::element_blank()
      )
  }

  if (!is.null(subtitle)) {
    fig.boxplot <- fig.boxplot +
      ggplot2::theme(
        plot.subtitle = ggplot2::element_text(size = 10))
  }

  if (facet.columns) {
    fig.boxplot <- fig.boxplot + ggplot2::facet_grid(GROUP ~ ., scales = "free")
    n.facet <- n.group * 2
    width <- 15
    height <- 5 + (4 * n.group)
  }

  if (facet.rows) {
    fig.boxplot <- fig.boxplot + ggplot2::facet_grid(FACET_ROWS ~ ., scales = "free")
    n.facet <- n.group * 2
    width <- 10
    height <- 5 + (4 * n.group)
  }

  if (!facet.rows && !facet.columns) {
    width <-  13 + (5 * n.group) + 1
    height <-  5
  }

  print(fig.boxplot)
  if (!is.null(bp.filename)) {
    suppressMessages(ggplot2::ggsave(
      filename = file.path(path.folder, bp.filename),
      plot = fig.boxplot,
      width = width,
      height = height,
      dpi = 300, units = "cm",
      limitsize = FALSE,
      useDingbats = FALSE
    )
    )
  }
  return(fig.boxplot)
}#Endboxplot_stats

[clauswilke]

Are you using or loading ggtern by any chance? The error message you quote does not exist in the ggplot2 code base anymore, in fact we don't even use stop() anymore.

The error you quote appears once in our revdep check, for the package tricolore, which uses ggtern:
https://github.com/tidyverse/ggplot2/blob/master/revdep/problems.md

As an aside, code such as the following doesn't make a lot of sense, because theme_bw() overwrites any previous theme settings.

  fig.boxplot <- fig.boxplot +
    ggplot2::theme(
      plot.title = ggplot2::element_text(size = 12, face = "bold", hjust = 0.5),
      legend.position = "none",
      axis.title.y = element.text,
      axis.text.y = element.text
    ) +
    ggplot2::theme_bw()

[thierrygosselin]

ggtern it is...

I do a requireNamespace for that package, because it's sometimes used at the end of that function (The one at the top level) to produce a plot, so if the argument is selected I check for that package. It explains why I was not getting the error when doing by-hand the other one...

Good catch!

as for the theme_bw() I know it looks counterintuitive... I was testing the difference if it's used before or after the ggplot2::theme and impact on background grid and text.

Thanks!

[clauswilke]

as for the theme_bw() I know it looks counterintuitive...

This is not about counterintuitive. It's that I know for an absolute fact that theme_bw() overwrites all prior theme settings. This may not have been fully in force before 3.3.0 but it is now, and if there's a case where it's not I'd consider it a bug.

[thierrygosselin]

Fig 1, 2 and 3 with theme_bw() in different places in the code....(after, not at all, or before the other ggplot2::theme). I was more interested in the different impact it as on the background...