Fix formals of compute_layer()

[yutannihilation]

Fixes #3202

Stat*$compute_layer() and Position*$compute_layer() have various names of formals, which seems for no valid reason; just because the common formals were so at that time (c.f. #3202 (comment)).

It seems these methods are not passed parameters by name, but by positions only, so I think this is safe to fix:

$ git grep "compute_layer("
R/layer.r:    self$stat$compute_layer(data, params, layout)
R/layer.r:    self$position$compute_layer(data, params, layout)
R/position-.r:#'   - `compute_layer(self, data, params, panel)` is called once
R/stat-.r:#'     `compute_layer(self, data, scales, ...)`,
R/stat-.r:#'     `compute_layer()` is called once per layer, `compute_panel_()`
R/stat-bindot.r:    ggproto_parent(Stat, self)$compute_layer(data, params, panels)
man/ggplot2-ggproto.Rd:\code{compute_layer(self, data, scales, ...)},
man/ggplot2-ggproto.Rd:\code{compute_layer()} is called once per layer, \code{compute_panel_()}
man/ggplot2-ggproto.Rd:\item \code{compute_layer(self, data, params, panel)} is called once

This PR fixes the inconsistency. I expect this would break nothing, but changing the argument names is (technically) a breaking change. So, I added a NEWS item for this.

[clauswilke]

I feel something is missing here. "have now the same names of arguments as other compute_layer() functions"?

[yutannihilation]

Thanks. I should probably say something like "The arguments of Stat*$compute_layer() and Position*$compute_layer() are now renamed to match the ones of Stat$compute_layer() and Position$compute_layer()." Does this make sense...?

[clauswilke]

Yes, maybe with this edit: "...now renamed to always match the ones...". After all, they mostly were matching already, but not always.

[yutannihilation]

Got it!

[thomasp85]

LGTM

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/