coord_radial: add xlim and ylim arguments

[Yunuuuu]

Partial of #6259 (The feature request has been changed from set limits for coord_radial()).

Since coord_polar() does not have an expand argument, I have not implemented this for coord_polar().

ggplot(mtcars, aes(disp, mpg)) +
        geom_point() +
        coord_radial(
            start = -0.4 * pi,
            end = 0.4 * pi, inner.radius = 0.3
        )

ggplot(mtcars, aes(disp, mpg)) +
        geom_point() +
        coord_radial(
            start = -0.4 * pi,
            end = 0.4 * pi, inner.radius = 0.3,
            xlim = c(200, 300),
            ylim = c(15, 30),
            clip = "on"
        )

[teunbrand]

Thanks for the PR! For reasons mentioned in #6259 (comment), I oppose this proposal, so I'll leave the task of approving this PR to other maintainers.

[Yunuuuu]

Thanks for your feedback! I understand your concerns and will leave this open for further review by other maintainers.

[teunbrand]

I discussed this with Thomas and he didn't object too much to this idea. However, we felt it was more intuitive to have a thetalim and rlim arguments for polar coordinates than xlim/ylim. Would you mind altering the parameterisation or do you have strong reasons why xlim/ylim would be preferred?

[Yunuuuu]

Thanks for the suggestions! I see the reasoning behind using thetalim and rlim for polar coordinates, as it makes the parameters more intuitive for their respective dimensions. My initial preference for xlim/ylim was to maintain consistency with existing coordinate systems in ggplot2, but I understand the value of explicitly distinguishing polar plot parameters. I'm happy to adjust the parameterization accordingly and will make the necessary changes tonight.

[teunbrand]

I think the above assumes theta == "x", whereas the limits should be swapped if theta == "y"

[Yunuuuu]

Should the expand arguments respect theta and r instead of x and y, as it currently does?

[teunbrand]

I don't think the top/right/bottom/left parameterisation of the expand argument makes much sense in the context of coord_radial(). You can rotate the plot with start/end and it is hard to say what will end up on which side, so we can leave it as is to avoid ruffling any feathers.

[Yunuuuu]

I've observed that even with clipping enabled, two points remain outside the circular panel area. I'm unsure whether grid package supports clipping with a custom polygon.

ggplot(mtcars, aes(disp, mpg)) +
        geom_point() +
        coord_radial(
            start = -0.4 * pi,
            end = 0.4 * pi, inner.radius = 0.3,
            xlim = c(200, 300),
            ylim = c(15, 30),
            clip = "on"
        )

[Yunuuuu]

It seems gridGeometry can help to do this: https://www.stat.auckland.ac.nz/~paul/Reports/Geometry/offset/offset.html

[teunbrand]

Using the current main branch, I don't see an issue with turning on clipping. I'm not sure why that doesn't work on your end.

devtools::load_all("~/packages/ggplot2/")
#> ℹ Loading ggplot2

ggplot(mtcars, aes(disp, mpg)) +
  geom_point() +
  scale_x_continuous(limits = c(200, 300), oob = scales::oob_keep) +
  scale_y_continuous(limits = c(15, 30), oob = scales::oob_keep) +
  coord_radial(
    start = -0.4 * pi,
    end = 0.4 * pi, inner.radius = 0.3,
    clip = "on"
  )

^{Created on 2025-04-03 with reprex v2.1.1}

[Yunuuuu]

Thanks for checking! Strangely, I still can't figure out why it's not working on my end — even with the current main branch.

[Yunuuuu]

devtools::load_all()
ggplot(mtcars, aes(disp, mpg)) +
    geom_point() +
    scale_x_continuous(limits = c(200, 300), oob = scales::oob_keep) +
    scale_y_continuous(limits = c(15, 30), oob = scales::oob_keep) +
    coord_radial(
        start = -0.4 * pi,
        end = 0.4 * pi, inner.radius = 0.3,
        clip = "on"
    )