get_layer_data(), get_layer_grob() can use layer names

[lgaborini]

A tiny PR that builds upon #5967.

Now the names can be used by get_layer_data() and get_layer_grob() in place of the layer index.
Only scalars are accepted, just like the current behavior.

If a name is supplied, it is used, but the layer must exist.
If both the name and the index are supplied (either explicitly or as a default argument), the name takes precedence.

The names must be unique (this is currently enforced by #5967 — should revisit whether this is no longer the case).

devtools::load_all(".")
p <- ggplot(head(mpg), aes(displ, hwy)) +
  geom_point(name = "foo") +
  geom_line(name = "bar")

names(p$layers)
#> [1] "foo" "bar"

get_layer_data(p, i = 2L)
#>     x  y PANEL group flipped_aes colour linewidth linetype alpha
#> 1 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 2 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 3 2.0 31     1    -1       FALSE  black       0.5        1    NA
#> 4 2.0 30     1    -1       FALSE  black       0.5        1    NA
#> 5 2.8 26     1    -1       FALSE  black       0.5        1    NA
#> 6 2.8 26     1    -1       FALSE  black       0.5        1    NA
get_layer_data(p, name = "bar")
#>     x  y PANEL group flipped_aes colour linewidth linetype alpha
#> 1 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 2 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 3 2.0 31     1    -1       FALSE  black       0.5        1    NA
#> 4 2.0 30     1    -1       FALSE  black       0.5        1    NA
#> 5 2.8 26     1    -1       FALSE  black       0.5        1    NA
#> 6 2.8 26     1    -1       FALSE  black       0.5        1    NA

# silently use the name
get_layer_data(p, i = 1L, name = "bar")
#>     x  y PANEL group flipped_aes colour linewidth linetype alpha
#> 1 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 2 1.8 29     1    -1       FALSE  black       0.5        1    NA
#> 3 2.0 31     1    -1       FALSE  black       0.5        1    NA
#> 4 2.0 30     1    -1       FALSE  black       0.5        1    NA
#> 5 2.8 26     1    -1       FALSE  black       0.5        1    NA
#> 6 2.8 26     1    -1       FALSE  black       0.5        1    NA

# no match by name: error

get_layer_data(p, name = "none")
#> Error in `get_layer_data()`:
#> ! `name` must be one of "foo" or "bar", not "none".

# no match by subscript: error (uglier)

get_layer_data(p, i = 4L)
#> Error in ggplot_build(plot)@data[[idx]]: subscript out of bounds

# Similarly for get_layer_grob()

get_layer_grob(p, i = 2L)
#> [[1]]
#> polyline[GRID.polyline.1]
get_layer_grob(p, name = "bar")
#> [[1]]
#> polyline[GRID.polyline.2]

Perhaps it's better to make i and name mutually exclusive, like id and number for recipes::tidy.recipe()?

[teunbrand]

Thank you for the PR! I agree that being able to use the name to get layer data/grob would be useful. I think it'd make more sense to merge to name approach in this PR with i. Using vctrs::vec_as_location2() to construct the index would handle both the integer and named cases.

[lgaborini]

Thanks, I didn't know that helper existed!

Now the parameter i is a bit inconsistent with get_panel_scales() since it indexes into panels instead of layers.
Probably not an issue until the panels get names too, or layers become panel-dependent.

[teunbrand]

Thanks for these updates, I like the direction of the PR!

Now the parameter i is a bit inconsistent with get_panel_scales() since it indexes into panels instead of layers.

I agree that it is unlikely to be an issue.

[teunbrand]

Thank you for the contribution and putting up with our comments! I've taken the liberty of implementing the test proposals and adding a news bullet.

[lgaborini]

Thanks, sorry but I had no time to finalize!

[teunbrand]

No worries, we're glad you contributed the main parts!