Partial polar plots

[teunbrand]

This is my take on #4462, that adds an end argument to coord_polar() to allow partial polar plots.

In brief:

• There are no visual changes if end is not used.
• It was straightforward to rescale to [start, end] instead of [0, 2π] to get the partial polar coordinates.
• After that rescaling, I had to account for the bounding box of the circle sector.
• The bit I found more difficult were the radius axis placement. I've resolved this as detailed below.

Normally, the axis is placed in the upper part of the y-axis. If we have only the lower part of a circle, I had to reflect the break positions:

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

p <- ggplot(mpg, aes(displ, hwy, colour = factor(cyl))) +
  geom_point()

p + coord_polar(start = 0.5 * pi, end = 1.5 * pi)

If 0 or π are not in the [start, end] interval, I have placed the axis in the x-direction. (Below also reflected because left instead of right).

p + coord_polar(start = 1.25 * pi, end = 1.75 * pi)

If no integer multiple of 0.5π is in that interval, I can't appropriately place an axis anywhere. In that case, I've thrown a message with instructions.

p + coord_polar(start = 0.1 * pi, end = 0.4 * pi)
#> Could not find appropriate placement for the radius axis.
#> ℹ A radius axis requires the [`start`-`end`] range to include one of: `c(0,
#>   0.5, 1, 1.5) * pi`.

^{Created on 2022-11-03 by the reprex package (v2.0.1)}

After the guides are ported over to the ggproto system, we might consider a more complete solution for axis placement than this.

[thomasp85]

Is this still meaningful to get in or should we tackle it from anew after ggproto guides have been merged?

[teunbrand]

I think the new ggproto guides don't interact with coord_polar() other than making coord_polar() skip doing proper guide stuff (as it wasn't using this), so these changes can safely co-exist. After the ggproto guides are merged, we can tackle bringing coord_polar() into the fold of guide-using coords, either in this PR or a new one. One might for example imagine an angle guide and radius guide for coord_polar().

[thomasp85]

ok, I'll do a review next week

[teunbrand]

Closing this PR in favour of #5334.