scale option that preserves scale but not location

[RoyalTS]

When facetting over things with different range, I currently have the option to give all facets the same axis limits, like so:

iris |> 
  group_by(Sepal.Width) |> 
  summarize(across(ends_with('Length'), mean)) |> 
  tidyr::pivot_longer(ends_with('Length')) |> 
  ggplot(aes(Sepal.Width, value, color=name)) +
  geom_line() +
  facet_wrap(~name)

or to free axes completely, like so:

iris |> 
  group_by(Sepal.Width) |> 
  summarize(across(ends_with('Length'), mean)) |> 
  tidyr::pivot_longer(ends_with('Length')) |> 
  ggplot(aes(Sepal.Width, value, color=name)) +
  geom_line() +
  facet_wrap(~name, scales='free_y')

free_* really does two things there though: It frees the scale of the axis – the pixel distance between the breaks – and it frees the absolute location of the scale – whether the break at 5 happens at the same horizontal point across facets.

I would love to have an option to scales that allows me to free location while maintaining scale. That is, vertical distances within each facet would correspond to the same absolute differences in the variable being plotted, but those differences might occur from another baseline.

An admittedly very hacky demonstration:

iris |> 
  group_by(Sepal.Width) |> 
  summarize(across(ends_with('Length'), mean)) |> 
  tidyr::pivot_longer(ends_with('Length')) -> plot_data

plot_data |>
  summarize(across(value, list(min=min, max=max)), .by=name) |> 
  mutate(
    range=value_max-value_min,
    midpoint = value_min + range/2
  ) |> 
  mutate(
    # center on the midpoint of the largest range among the facets
    max_range = max(range),
    axis_min = midpoint - max_range / 2,
    axis_max = midpoint + max_range / 2
  ) -> facet_spec

make_facet <- function(df, x_var, axis_min, axis_max) {
  df |>
    filter(name == x_var) |> 
    ggplot(aes(Sepal.Width, value)) +
    geom_line() +
    facet_wrap(~name) +
    coord_cartesian(ylim=c(axis_min, axis_max))
}

facet_spec |> 
  rowwise() |> 
  mutate(plots = list(make_facet(plot_data, name, axis_min,	axis_max))) |> 
  pull(plots) |> 
  patchwork::wrap_plots()

[teunbrand]

If you pre-calculate the maximum range per facet, you can just use a function as the limits argument that will set a consistent range. I'm unsure whether this is widely applied enough to merit an additional option in facets.

library(tidyverse)

df <- iris |> 
  group_by(Sepal.Width) |> 
  summarize(across(ends_with('Length'), mean)) |> 
  pivot_longer(ends_with('Length'))

max_range <- df |>
  group_by(name) |>
  summarise(range = diff(range(value))) |>
  summarise(range = max(range)) |>
  pull()

ggplot(df, aes(Sepal.Width, value, color=name)) +
  geom_line() +
  facet_wrap(~name, scales = "free_y") +
  scale_y_continuous(
    limits = \(x) mean(x) + c(-0.5, 0.5) * max_range
  )

^{Created on 2023-08-02 with reprex v2.0.2}

[RoyalTS]

Oh! I had no idea limits could take a function as an argument! That'll do!