element_geom use of linewidth & borderwidth etc

[davidhodge931]

Looking at element_geom, I noticed a couple of things:

1. linewidth/linetype and borderwidth/borderwidth doesn't seem internally consistent.. See example code below. Not sure how to fix this..
2. Likewise, I think it'd be more intuitive if argument names in element_geom names would map to aesthethetics as much as possible. So I think pointsize sould instead be size etc with a separate element_geom_font function.

library(tidyverse)

p <- ggplot(economics, aes(date)) +
  geom_density()

#would expect this to work
ggplot(economics, aes(date)) +
  geom_density() +
  theme(
    geom.density = element_geom(
      linewidth = 5,
      linetype = 3
    )
  )

#given you can do this
ggplot(economics, aes(date)) +
  geom_density(
    linewidth = 5,
    linetype = 3
  )

#instead have to do this
ggplot(economics, aes(date)) +
  geom_density() +
  theme(
    geom.density = element_geom(
      borderwidth = 5,
      bordertype = 3
    )
  )

#but you can't do this
ggplot(economics, aes(date)) +
  geom_density(
    borderwidth = 5,
    bordertype = 3
  )
#> Warning in geom_density(borderwidth = 5, bordertype = 3): Ignoring unknown
#> parameters: `borderwidth` and `bordertype`

^{Created on 2025-06-17 with reprex v2.1.1}

[davidhodge931]

3. I also appreciate what you guys are trying to do with ink/paper/accent, but to me it doesn't look worth it in terms of the confusion with the API that it creates

I would way prefer it was just like this as much as possible

[thomasp85]

We decided that as far as geom default go, there's a difference between lines and strokes – you may eg want a thinner default stroke than line. Now for geom_density in particular this is a bit surprising because it is a polygon under the hood so it present itself as a line but is in fact a polygon. Maybe we can do something there.

As for the last point, I don't consider the internals of ggproto classes part of the API of ggplot2 and have no issue making them appear a bit more complicated in the service of a valuable feature

[teunbrand]

Yeah I agree with Thomas here. There are different roles that appear between geoms but not really within geoms, which in my mind justifies the deviation.

[davidhodge931]

For the font size arguments in element_geom, could you remove these and rename pointsize and pointshape to size and shape.

Then if people want to change the size of points and text, they can just add:

  theme(
    geom = element_geom(size = 2),
    geom.text = element_geom(size =10),
    geom.label = element_geom(size =10),
  ) 

The borderwidth and bordertype seems useful only for setting everything with geom = element_geom(). But people can still write out the border ones as they need them.

  theme(
    geom = element_geom(linewidth = 1),
    geom.density = element_geom(linewidth = 0.5),
    geom.bar = element_geom(linewidth = 0.5),
    geom.polygon = element_geom(linewidth = 0.5),
  ) 

As you can use theme_set, you only have to do all this once.

If you don't want to do any of that, an alternative way with borderwidth/bordertype is to make it inherit from linewidth/linetype - and likewise bordertype? That way if someone gets it wrong, it still works as they intend it to (e.g. as example in density)

I'd put the ink, paper, accent features as the last args that advanced people could use if they want to.

My 2 cents anyway. Thanks for all the ggplot2 work

[davidhodge931]

Here's how the args in the function would look after removing the font specific args, and making borderwidth and bordertype inherit. As noted above, people can add separate geom.text and geom.label layers to adjust the size of the font. And there is always an option to create a element_geom_font now or in future. I think these changes would help with the intuitiveness of the API

element_geom(
  colour = NULL,
  color = NULL,
  fill = NULL
  size = NULL,
  shape = NULL,
  family = NULL,
  linewidth = NULL,
  linetype = NULL,
  borderwidth = linewidth,
  bordertype = linetype,
  ink = NULL,
  paper = NULL,
  accent = NULL
)

[davidhodge931]

The borderwidth and bordertype arguments are pretty confusing, relative to the benefit they add.. and that's not even getting into whether boxplot is borderwidth or linewidth

library(tidyverse)

#add borderwidth to affect linewidth
set_theme(theme_grey() +
            theme(geom.density = element_geom(borderwidth = 4)))

get_geom_defaults("density")
#>   colour fill weight alpha linewidth linetype
#> 1  black   NA      1    NA         4        1

#but linewidth does not affect linewidth??
set_theme(theme_grey() +
            theme(geom.density = element_geom(linewidth = 4)))

get_geom_defaults("density")
#>   colour fill weight alpha linewidth linetype
#> 1  black   NA      1    NA       0.5        1

#and borderwidth does not affect linewidth here??
ggplot(economics, aes(date)) +
  geom_density(borderwidth = 2)
#> Warning in geom_density(borderwidth = 2): Ignoring unknown parameters:
#> `borderwidth`

^{Created on 2025-06-19 with reprex v2.1.1}

[teunbrand]

I'd put the ink, paper, accent features as the last args that advanced people could use if they want to.

I disagree with this. ink and paper are the first things people should try to tinker with if they're using element_geom().

For the font size arguments in element_geom, could you remove these and rename pointsize and pointshape to size and shape.

I don't like this. I want absolutely zero geom.foo hierarchy in base ggplot2 themes to minimise the amount of surprise that some current theme elements create (like why is axis text not red when you use theme_gray() + theme(text = element_text(colour = "red"))).

borderwidth = linewidth,
bordertype = linetype,

I like this idea in principle, but it may get confusing when you just want to update the linewidth without also affecting the borderwidth (or types).