[R-pkg-devel] Removing import(methods) stops exporting S4 "meta name"

[Duncan Murdoch]

To save others a bit of time, the package in question is "do", and 
Michael has already posted an issue on their Github page asking why they 
made this import:  https://github.com/yikeshu0611/do/issues/1 .

And if you remove that import and run R CMD check on "do", it fails with 
this error:

* checking examples ... ERROR
Running examples in ‘do-Ex.R’ failed
The error most likely occurred in:

 > ### Name: join
 > ### Title: Join two dataframes together
 > ### Aliases: join join_inner join_full join_left join_right join_out
 >
 > ### ** Examples
 >
 > df1=data.frame(x=rep(c('b','a','c'),each=3),
+               y=c(1,3,6),
+               v=1:9)
 >
 > df2=data.frame(x=c('c','b','e'),
+                v=8:6,
+                foo=c(4,2,1))
 > join_inner(df1,df2,'x')
Error in xj[i] : invalid subscript type 'list'
Calls: join_inner -> data.frame -> [ -> [.data.table -> [.data.frame
Execution halted

I'm pretty sure that import should not be made, but I don't know the 
intention of this example, or what the correct fix would be.

Duncan Murdoch

On 14/03/2024 2:34 p.m., Michael Chirico wrote:
> In an effort to streamline our NAMESPACE, we moved from blanket
> 'import(methods)' to specific importFrom(methods, ....) for the
> objects we specifically needed.
> 
> Doing so broke a downstream package, however, which has this directive:
> 
> importFrom(data.table,`.__T__[:base`)
> 
> That package stopped installing after the above change. I can get it
> to install again by adding:
> 
> importClassesFrom(methods, "[")
> 
> to our package, but I'm not sure why this would be necessary. I'm
> still left with two questions:
> 
>   1. How did we end up with ".__T__[:base" in our exports when we don't
> do any S4 around '[' in the package (we do export S3 methods on it)?
>   2. Is there any legitimate reason for a package to try and import
> such an object? In other words, is breaking this downstream package by
> not adding the 'importClassesFrom' workaround the right thing to do? I
> don't see any other CRAN packages with a similar directive in its
> NAMESPACE.
> 
> Michael Chirico
> 
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel