Feature request: Reasonable Representation of Spatial Data Axis Ranges within Faceting Processes

[Breeze-Hu]

Hello!
Facet plots in ggplot2 using spatial coordinates (coord_sf) will force the sharing of x and y axis ranges(?). facet_wrap and facet_grid both behave the same case.

I suspect this issue is complicated due to the potential existence of a fixed aspect ratio in coord_sf. This also makes it extremely difficult to use patchwork to stitch together multiple spatial results.

This will result in a very strange spatial data layout:

library(tidyverse)
library(rnaturalearthdata)
library(rnaturalearth)
#> 
#> Attaching package: 'rnaturalearth'
#> The following object is masked from 'package:rnaturalearthdata':
#> 
#>     countries110


world <- ne_countries()
#plot(world$geometry)

data <- world |>
  filter(name %in% c("Japan", "Austria", "Australia"))

d1 <- data |> mutate(value = 1)
d2 <- data |> mutate(value = 2)
d3 <- data |> mutate(value = 3)

dt <- rbind(d1, d2, d3)

ggplot() +
  geom_sf(data = dt, aes(fill = value), color = "black") +
  scale_fill_viridis_c() +
  facet_grid(value ~ name) +
  theme_minimal()

ggplot() +
  geom_sf(data = dt, aes(fill = value), color = "black") +
  scale_fill_viridis_c() +
  facet_wrap(c("value", "name")) +
  theme_minimal()

ggplot() +
  geom_sf(data = dt, aes(fill = value), color = "black") +
  scale_fill_viridis_c() +
  facet_wrap(c("value", "name"), scales = 'free_x') +
  theme_minimal()
> Error in `draw_panels()`:
! `facet_wrap()` can't use free scales with `coord_sf()`.
Run `rlang::last_trace()` to see where the error occurred.

^{Created on 2025-10-01 with reprex v2.1.1}

Best wishes,
Hu

[teunbrand]

Some related discussion occurred in #2651.

In particular, from #2651 (comment):

To implement what you're requesting, one would have to write a training algorithm that trains x and y axes simultaneously while producing a targeted aspect ratio. This is possible in principle, but it's not immediately obvious how to do it in the current code base, at least to me. The best way forward might be to write a new faceting function specifically for this purpose, probably first in the form of an extension package.

Which declares this out-of-scope for ggplot2 itself. It would also not be immediately obvious to me how to do this, so for practical reasons I'm inclinded to agree.

[Breeze-Hu]

Thanks!

I agree this is a complex situation. as a reference for people later, tmap might be an option to pursue for ideal results tmap_facets-one-dimensional. but this also means giving up ggplot2's extensive ecosystem...

If you decide now is not the appropriate time to continue discussing this issue, feel free to close it. 🤣

[teunbrand]

The alternative to giving up ggplot2's ecosystem for tmap, is to extend ggplot2's ecosystem to include the same functionality that tmap offers :)

[Breeze-Hu]

Hi!
While continuing to experiment with this issue, I encountered something interesting:
I imagined that when the aspect ratio is fixed, changing either the length or width should automatically adjust the other. However, the warning indicated that the aspect ratio had changed (and indeed the result did), while the length remained unchanged. Is this intentional? 🤣

ggplot() +
  geom_sf(
    data = subset(dt, dt$name == "Japan"),
    aes(fill = value),
    color = "black"
  ) +
  scale_fill_viridis_c() +
  facet_wrap(c("value"), nrow = 3) +
  theme_minimal()+
  theme(panel.widths = unit(3, 'cm'))  # Adjusting the width only

Warning message:
Aspect ratios are overruled by panel.widths theme element.

[teunbrand]

Yes this is intentional. The warning message covers the exact surprise you're confronted with, which means we anticipated it when we wrote the message :)

[Breeze-Hu]

If the user doesn't want to change the aspect ratio but only adjust the size, this task is intentionally left to the final output rendering stage (ggsave()). Is my understanding correct?
: )

[teunbrand]

Only setting one of the dimensions with theme(panel.widths, panel.heights) will cause the other to behave as usual: it will take up the available space afforded to it by the graphics device.

[Breeze-Hu]

Thanks!
Is there a way to adjust the height and width while keeping the aspect ratio unchanged? This would facilitate potential future patchwork to stitch multiple plots together. Alternatively, could users get the aspect ratio and manually control it via the theme() function?

Thank you for your patience! I really have so many things I don't understand. 🤣

[teunbrand]

Is there a way to adjust the height and width while keeping the aspect ratio unchanged?

Only if you know the aspect ratio in advance, in which case you can set the size via the theme accordingly.

Alternatively, could users get the aspect ratio and manually control it via the theme() function?

I think you'd have to render a plot once without a fixed size, rummage through the internal structures to excavate the apsect ratio, and then render the plot again with a fixed size.

I'll look into if we can't 'infer' the other dimension when the aspect ratio is known.

[teunbrand]

OK, I recall now why we don't preserve aspect ratio.
Here, we give different width instructions to different panels.
If we maintain the aspect ratio, the results should be 2cm and 8cm high.
We wouldn't be able to align the result.

library(ggplot2)

ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  facet_wrap(~ year) +
  theme(
    aspect.ratio = 2,
    panel.widths = unit(c(2, 4), "cm")
  )
#> Warning: Aspect ratios are overruled by `panel.widths` theme element.

^{Created on 2025-10-01 with reprex v2.1.1}

[Breeze-Hu]

Ohhh, yes. I see.
It must be said that there are some differences between geospatial mapping and statistical graphics. This is an objective reality.

Thanks so much！👍👍

[teunbrand]

feel free to close it. 🤣

I think I'll take you up on this