ggplot2 add minor tick marks outside plotting area without turning clip off

Question

I'm looking for a way to add minor tick marks to ggplots without using coord_cartesian(clip = "off"). Or a way to reproducible apply clipping to the x axis but not the y axis, or vice versa.

So far I have been using the annotation_ticks() function and the GeomTicks geom defined in this excellent answer here (with some minor modifications to make it work with ggplot2 v3.3.0). Unfortunately, for the tick mark annotations to appear on the outside of the plot one must use coord_cartesian(clip = "off") which means anything else that sits outside the plotting area is also exposed (see reprex below).

Alternatively, perhaps there is a way to leverage any of the new capabilities of ggplot2 v3.3.0 to draw minor ticks not as an annotation but as an actual part of the axis/plot so that it is possible to draw them outside the plotting area.

I am not a software developer but perhaps one could define a new theme element using register_theme_elements called axis.minor.ticks that behaves like axis.ticks but gets the appropriate location of the minor ticks from panel_params$y$break_positions_minor instead of panel_params$y$break_positions. Or somehow use the new guide_x() S3 functions.

Any help would be much appreciated!

Annotation function and ggproto object

The annotation_ticks() function (incorporating this fix for faceting issue):

annotation_ticks <- function(sides = "b",
                             scale = "identity",
                             scaled = TRUE,
                             ticklength = unit(0.1, "cm"),
                             colour = "black",
                             size = 0.5,
                             linetype = 1,
                             alpha = 1,
                             color = NULL,
                             ticks_per_base = NULL,
                             data = data.frame(x = NA), 
                             ...) {
  if (!is.null(color)) {
    colour <- color
  }

  # check for invalid side
  if (grepl("[^btlr]", sides)) {
    stop(gsub("[btlr]", "", sides), " is not a valid side: b,t,l,r are valid")
  }

  # split sides to character vector
  sides <- strsplit(sides, "")[[1]]

  if (length(sides) != length(scale)) {
    if (length(scale) == 1) {
      scale <- rep(scale, length(sides))
    } else {
      stop("Number of scales does not match the number of sides")
    }
  }

  base <- sapply(scale, function(x) switch(x, "identity" = 10, "log10" = 10, "log" = exp(1)), USE.NAMES = FALSE)

  if (missing(ticks_per_base)) {
    ticks_per_base <- base - 1
  } else {
    if ((length(sides) != length(ticks_per_base))) {
      if (length(ticks_per_base) == 1) {
        ticks_per_base <- rep(ticks_per_base, length(sides))
      } else {
        stop("Number of ticks_per_base does not match the number of sides")
      }
    }
  }

  delog <- scale %in% "identity"

  layer(
    data = data,
    mapping = NULL,
    stat = StatIdentity,
    geom = GeomTicks,
    position = PositionIdentity,
    show.legend = FALSE,
    inherit.aes = FALSE,
    params = list(
      base = base,
      sides = sides,
      scaled = scaled,
      ticklength = ticklength,
      colour = colour,
      size = size,
      linetype = linetype,
      alpha = alpha,
      ticks_per_base = ticks_per_base,
      delog = delog,
      ...
    )
  )
}

The ggproto object (now works with ggplot2 v3.3.0):

GeomTicks <- ggproto(
  "GeomTicks", Geom,
  extra_params = "",
  handle_na = function(data, params) {
    data
  },

  draw_panel = function(data,
                        panel_scales,
                        coord,
                        base = c(10, 10),
                        sides = c("b", "l"),
                        scaled = TRUE,
                        ticklength = unit(0.1, "cm"),
                        ticks_per_base = base - 1,
                        delog = c(x = TRUE, y = TRUE)) {
    ticks <- list()

    for (s in 1:length(sides)) {
      if (grepl("[b|t]", sides[s])) {

        # for ggplot2 < 3.3.0 use: xticks <- panel_params$x.minor
        if (utils::packageVersion("ggplot2") >= "3.2.1.9000") {
          x_minor_breaks <- panel_scales$x$break_positions_minor()
          x_major_breaks <- panel_scales$x$break_positions()
        } else {
          x_minor_breaks <- panel_scales$x.minor
          x_major_breaks <- panel_scales$x.major
        }

        xticks <- setdiff(x_minor_breaks, x_major_breaks)

        # Make the grobs
        if (grepl("b", sides[s])) {
          ticks$x_b <- with(
            data,
            segmentsGrob(
              x0 = unit(xticks, "npc"),
              x1 = unit(xticks, "npc"),
              y0 = unit(0, "npc"),
              y1 = ticklength,
              gp = gpar(
                col = alpha(colour, alpha),
                lty = linetype,
                lwd = size * .pt
              )
            )
          )
        }
        if (grepl("t", sides[s])) {
          ticks$x_t <- with(
            data,
            segmentsGrob(
              x0 = unit(xticks, "npc"),
              x1 = unit(xticks, "npc"),
              y0 = unit(1, "npc"),
              y1 = unit(1, "npc") - ticklength,
              gp = gpar(
                col = alpha(colour, alpha),
                lty = linetype,
                lwd = size * .pt
              )
            )
          )
        }
      }


      if (grepl("[l|r]", sides[s])) {

        # for ggplot2 < 3.3.0 use: yticks <- panel_params$y.minor
        if (utils::packageVersion("ggplot2") >= "3.2.1.9000") {
          y_minor_breaks <- panel_scales$y$break_positions_minor()
          y_major_breaks <- panel_scales$y$break_positions()
        } else {
          y_minor_breaks <- panel_scales$y.minor
          y_major_breaks <- panel_scales$y.major
        }

        yticks <- setdiff(y_minor_breaks, y_major_breaks)

        # Make the grobs
        if (grepl("l", sides[s])) {
          ticks$y_l <- with(
            data,
            segmentsGrob(
              y0 = unit(yticks, "npc"),
              y1 = unit(yticks, "npc"),
              x0 = unit(0, "npc"),
              x1 = ticklength,
              gp = gpar(
                col = alpha(colour, alpha),
                lty = linetype, lwd = size * .pt
              )
            )
          )
        }
        if (grepl("r", sides[s])) {
          ticks$y_r <- with(
            data,
            segmentsGrob(
              y0 = unit(yticks, "npc"),
              y1 = unit(yticks, "npc"),
              x0 = unit(1, "npc"),
              x1 = unit(1, "npc") - ticklength,
              gp = gpar(
                col = alpha(colour, alpha),
                lty = linetype,
                lwd = size * .pt
              )
            )
          )
        }
      }
    }
    gTree(children = do.call("gList", ticks))
  },
  default_aes = aes(colour = "black", size = 0.5, linetype = 1, alpha = 1)
)

Graph with `coord_cartesian(clip = "on")`

Column with very thick line width looks fine but can't see tick annotations.

library(ggplot2)
library(grid)

ggplot(mpg, aes(x = class, y = displ, fill = class)) + 
  stat_summary(fun = mean, geom = "col", colour = "black", size = 1) + 
  theme_classic(base_size = 8) + 
  scale_y_continuous(limits = c(0, 8), expand = c(0, 0)) + 
  annotation_ticks(sides = "l", ticklength = -1 * unit(0.2, "cm")) + 
  coord_cartesian(clip = "on")

ggsave("clip_on.png", device = "png", width = 4, height = 3)

column plot with clip=on

Graph with `coord_cartesian(clip = "off")`

Tick annotations visible but columns with very thick line width are showing outside of the plotting area.

ggplot(mpg, aes(x = class, y = displ, fill = class)) + 
  stat_summary(fun = mean, geom = "col", colour = "black", size = 1) + 
  theme_classic(base_size = 8) + 
  scale_y_continuous(limits = c(0, 8), expand = c(0, 0)) + 
  annotation_ticks(sides = "l", ticklength = -1 * unit(0.2, "cm")) + 
  coord_cartesian(clip = "off")

ggsave("clip_off.png", device = "png", width = 4, height = 3)

column plot with clip=off

score 2 · Accepted Answer · answered Apr 07 '20 at 20:04

This code seemed earily familiar to me, so I'd thought to weigh in.

Yes, with ggplot v3.3.0 guides have become extendible, though I doubt they'll be in their current form for a long time because through the grapevines I've heard they want to switch guides to the ggproto system too.

The cheapest way without too many bells and whisles to do what you ask, is to adjust the guide training portion of guides. Since this is an S3 method, we'll need a new guide class to write a custom method:

library(ggplot2)
library(rlang)
#> Warning: package 'rlang' was built under R version 3.6.3
library(glue)

guide_axis_minor <- function(
  title = waiver(), check.overlap = FALSE, angle = NULL,
  n.dodge = 1, order = 0, position = waiver()
) {
  structure(list(title = title, check.overlap = check.overlap, 
                 angle = angle, n.dodge = n.dodge, order = order, position = position, 
                 available_aes = c("x", "y"), name = "axis"), 
            class = c("guide", "axis_minor", "axis"))
}

You'll note that the function above is identical to guide_axis(), except for an extra class. The order of classes is important here, because we're subclassing the axis class, so that we can be lazy and just use all the methods that already exist.

This brings us to training, truly the only thing that needs to be adjusted a bit. I've commented in the relevant bits. The majority of the function is still identical to guide_train.axis internal function. Briefly, we're treating minor breaks as major breaks with empty labels.

guide_train.axis_minor <- function(guide, scale, aesthetic = NULL) {
  aesthetic <- aesthetic %||% scale$aesthetics[1]

  # Seperately define major and minor breaks
  major_breaks <- scale$get_breaks()
  minor_breaks <- scale$get_breaks_minor()

  # We set the actual breaks to be both major and minor
  breaks <- union(major_breaks, minor_breaks)

  # We keep track of what breaks were the major breaks
  is_major <- breaks %in% major_breaks

  empty_ticks <- ggplot2:::new_data_frame(
    list(aesthetic = numeric(), .value = numeric(0), .label = character())
  )
  if (length(intersect(scale$aesthetics, guide$available_aes)) == 0) {
    warn(glue("axis guide needs appropriate scales: ", 
              glue_collapse(guide$available_aes, ", ", last = " or ")))
    guide$key <- empty_ticks
  } else if (length(breaks) == 0) {
    guide$key <- empty_ticks
  } else {
    mapped_breaks <- if (scale$is_discrete()) {
      scale$map(breaks)
    } else {
      breaks
    }
    ticks <- ggplot2:::new_data_frame(setNames(list(mapped_breaks), 
                                     aesthetic))
    ticks$.value <- breaks
    ticks$.label <- scale$get_labels(breaks)

    # Now this is the bit where we set minor breaks to have empty labls
    ticks$.label[!is_major] <- ""

    guide$key <- ticks[is.finite(ticks[[aesthetic]]), ]
  }
  guide$name <- paste0(guide$name, "_", aesthetic)
  guide$hash <- digest::digest(list(guide$title, guide$key$.value, 
                                    guide$key$.label, guide$name))
  guide
}

Then, because we subclassed the axis class, all the functions written for that class will also work for our axis_minor class, so we're done. Now you can just call the guide from any continuous position scale by name:

ggplot(mpg, aes(x = class, y = displ, fill = class)) + 
  stat_summary(fun = mean, geom = "col") + 
  scale_y_continuous(limits = c(0, 8), 
                     guide = "axis_minor")

^{Created on 2020-04-07 by the reprex package (v0.3.0)}

Thanks heaps! This solves the question as posed above. Is there any way to make the minor ticks shorter than the major ticks? As they are both `ticks` they are both affected by `axis.ticks.length`. Perhaps by modifying the `ticks_grob` in `draw_axis()`? — csdaw, Apr 08 '20 at 14:08
That would indeed be the way to go. The laziest way to implement this, is write a `guide_gengrob` method for the new guide that wraps the `guide_gengrob.axis` method (which in turn wraps the `draw_axis()`) and make some final adjustments to the grob at the end. If you want to properly, you should probably add the `is_major` variable from the training to guide object and make a custom theme setting for the minor ticks. I think it is a bit too involved for a SO answer to do this and perhaps it would be best if that code would live somewhere on github. — teunbrand, Apr 08 '20 at 16:03

ggplot2 add minor tick marks outside plotting area without turning clip off

Annotation function and ggproto object

Graph with coord_cartesian(clip = "on")

Graph with coord_cartesian(clip = "off")

1 Answers1

Graph with `coord_cartesian(clip = "on")`

Graph with `coord_cartesian(clip = "off")`