Guides for coord_sf()

[teunbrand]

This PR aims to fix a problem akin to #3959, but for coord_sf().

Briefly, coord_sf() was doing its own guide thing, ignoring any user-provided guides.
Now, it renders the guides through the guides system.

Less briefly:
To make this work, I had to cast the graticule positions into a viewscale that the guides can use for training.
I was a bit surprised that axis placement is tightly controlled in coord_sf() itself, in that scale_x_continuous(position = "top") is ignored and just adding guides(x.sec = "axis") doesn't just set a top axis.
Maybe it would be good to allow some flexibility here, but I didn't change that behaviour in this PR.
I'm new-ish to {sf}, so if anybody knows of hairy edge cases I mightn't have thought of, please let me know.

To illustrate:

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

nc <- sf::st_read(system.file("shape/nc.shp", package = "sf"), quiet = TRUE)
ggplot(nc) +
  geom_sf() +
  coord_sf(crs = 3347, label_graticule = c("NSWE")) + # all labels
  guides(
    x = guide_none("guide_none() passes on their title"),
    y = guide_axis("angles work", angle = 45),
    x.sec = guide_axis("dodging works", n.dodge = 2),
    y.sec = guide_axis(angle = -45, n.dodge = 2)
  )

^{Created on 2023-05-01 with reprex v2.0.2}

[thomasp85]

@clauswilke you have spend the most time with sf - do you have the bandwidth to look at this?

[clauswilke]

@thomasp85 Yes, this touches mostly code that I wrote, so I guess I'll have to look at it carefully. I don't remember much of it, but I'll take a look.

@teunbrand Two quick items from my end:

1. Are we using Map() in ggplot? I thought we're generally using apply functions.
2. It looks to me like a bunch of comments have disappeared in porting the code. I would encourage you to comment a bit more generously. This is confusing code, and without comments nobody will understand what it does five years into the future.

[clauswilke]

To be clear: My code already didn't have that many comments. The new code has even fewer. I think the goal should be more comments than my original code had.

[teunbrand]

Are we using Map() in ggplot? I thought we're generally using apply functions.

Under the hood, Map() uses the apply-family of functions (mapply) but doesn't simplify output. Should there be a reason not to use it?

I think the goal should be more comments than my original code had.

Yeah this is a fair point and had some problem deciphering which parts were doing what as well.

[clauswilke]

Regarding Map(), I'll leave it to @thomasp85 to make this decision. As a general principle, since we're writing ggplot2 using somewhat obscure (for people coming from the tidyverse) base-R functionality, it makes sense to use a limited set of functions to keep the code easier to read. At least I always thought that we're doing this. I could be wrong.

[teunbrand]

I've sprinkled around some comments and improved some variable names. I'm at such familiarity with GIS that I have to google every time what a meridian and parallel is, so bare with me if I get them confused at times.

[clauswilke]

I've reviewed it and had two more minor comments. Overall this seems fine to me, though I will say I don't fully understand the guide system. In the long run, this will likely be more maintainable, though, so moving forward in this direction is good.

[teunbrand]

Thanks for the review Claus! Yeah I suppose the guide system will take some getting used to, but I'd be happy to try to answer any questions you might have.