Review error messages

[olivroy]

Add full stops to error messages + other tweaks. (to adhere to tidyverse style guide)

Also use expect_snapshot() over expect_snapshot_warning() and error_snapshot_error() for consistency. (they are superseded in testthat. It allows to see more clearly what error message the user sees.

I may do a transition to expect_snapshot() in a future PR. For the sake of having a self contained PR, I just reviewed error messages.

[teunbrand]

Thanks for such a thorough review of all the error messages! The PR looks almost good to me, but if you could have a look at some of the comments that'd be swell.

[olivroy]

I applied your suggestions. Thanks for the prompt review!

On a related note, I was wondering how to make error messages look better:

devtools::load_all("~/rrr-forks/ggplot2/")
#> ℹ Loading ggplot2
#> Warning: le package 'testthat' a été compilé avec la version R 4.3.2
plot <- ggplot(sf::st_polygon(list(matrix(1e3*c(1, 2, 3, 1, 1, 3, 2, 1), ncol = 2)))) +
  geom_sf()
p <- plot +
  scale_x_continuous(
    breaks = c(1000, 2000, 3000),
    labels = function(...) c("A", "B")
  )
ggplot_build(p)
#> Error in `fixup_graticule_labels()`
#> ! `breaks` and `labels` along `x` direction have different lengths.

^{Created on 2023-12-04 with reprex v2.0.2}

Like, how could it show this instead?

#> Error in `scale_x_continuous()`
#> ! `breaks` and `labels` along `x` direction have different lengths.

[teunbrand]

I think you could 'lie' and use:

cli::cli_abort({error_message}, call = scale_x$call)

I call this a 'lie' because it is some coord-specific quirck that doesn't let the scale handle breaks/labels as per usual, but the scale is definitely the place where it should be fixed.

[olivroy]

Ok thanks. I may look into this when more PRs are merged.

[teunbrand]

Look great, thanks for tackling these messages!