`ERROR: Normalized names are no longer unique` on `@rlibrary ggplot2`

[josswright]

Apologies if I'm making a very naive mistake!

I've been trying to start a long-planned push to begin migrating from R to Julia. One of the things that has kept me from trying earlier is the strength of ggplot2. When I saw that RCall might let me keep using ggplot2 for a while as I ease into Julia, it seemed perfect.

Running Julia version 1.12.4+0.x64.linux.gnu, installed via juliaup under Arch Linux, I ran into an immediate error with the basic examples I found online:

using RCall
@rlibrary ggplot2

... immediately produces:

ERROR: Normalized names are no longer unique: is_facet, is_ggplot, is_ggproto and is_theme
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:44
 [2] rimport(pkg::String, s::Symbol; normalizenames::Bool, normalization::Vector{Pair{Char, Char}})
   @ RCall ~/.julia/packages/RCall/JIY9I/src/namespaces.jl:38
 [3] rimport (repeats 2 times)
   @ ~/.julia/packages/RCall/JIY9I/src/namespaces.jl:23 [inlined]
 [4] rimport(::Symbol; kwargs::@Kwargs{})
   @ RCall ~/.julia/packages/RCall/JIY9I/src/namespaces.jl:55
 [5] top-level scope
   @ REPL[7]:1
 [6] top-level scope
   @ REPL[7]:1

I haven't seen this problem mentioned elsewhere with respect to ggplot2, except maybe some issues around automatic conversion of dots in function names to underscores from 2018: #234.

I assume I must be doing something wrong, but this is a completely fresh install of Julia where I've added only the RCall package.

I've not run into a problem with any other libraries I've tried: @rlibrary tidyverse doesn't seem to have any problems, for example.

Any help appreciated.

[josswright]

I've spent the evening playing with things to try and be sure that this wasn't just a new user error.

As far as I can tell, it does seem to be the obvious fact that is_facet, is_ggplot, is_ggproto, and is_theme in ggplot2 all have dotted equivalents: is.facet, etc. (Which are deprecated, but still need to be handled in the general sense.)

In this case they seem to be aliases, but this seems a fairly significant problem.

[josswright]

Digging into the source here:

RCall.jl/src/namespaces.jl

Lines 108 to 126 in e74538e

	"""
	@rlibrary(package)
	Load an R package as if it were a Julia package, bringing all exported bindings into scope.
	More or less equivalent to:
	```
	__temp__ = rimport("ggplot2")
	using .__temp__
	```
	"""
	macro rlibrary(x)
	tmp = gensym("RCall")
	ex = Expr(:block, __source__,
	Expr(:(=), esc(tmp),
	Expr(:call, GlobalRef(@__MODULE__, :rimport), QuoteNode(x))),
	Expr(:using, Expr(:., :., tmp)))
	return Expr(:toplevel, ex)
	end

... it looks like I can get around this by using rimport directly and overriding the default normalization, rather than using the @rlibrary macro:

__temp__ = rimport("ggplot2", normalization=['.' => '£'])
using .__temp__

(Using '£' here, because it seems that a single character is needed, and that's relatively unlikely to be conflicting with anything that could cause a problem.)

Might be good to document this somewhere in case anyone else runs into this problem?