R devel - Nov 2020 - Named class vector

The source to the noquote() function looks like this:

noquote <- function(obj, right = FALSE) {
     ## constructor for a useful "minor" class
     if(!inherits(obj,"noquote"))
         class(obj) <- c(attr(obj, "class"),
                         if(right) c(right = "noquote") else
"noquote")
     obj
}

Notice what happens with right = TRUE:

 > x <- noquote("a", right = TRUE)
 > x
[1] a
 > class(x)
     right
"noquote"

The class vector for x is named.  The print method pays attention to the 
name, so we get different behaviour for a class of "noquote" and a
class
of c(right = "noquote").

I had never noticed a named class vector before, and it raised some 
questions for me:

- Is this used anywhere else?
- Are names preserved in all the operations normally done on a class 
vector?  (As far as I can see they are, but maybe I've missed something.)
- Is it a good idea to encode a string value worth of information in the 
name, rather than setting the class to something like c("noquote", 
"right") instead?

Comments would be welcome.

Duncan Murdoch

>>>>> Duncan Murdoch writes:
> The source to the noquote() function looks like this:
> noquote <- function(obj, right = FALSE) {
>      ## constructor for a useful "minor" class
>      if(!inherits(obj,"noquote"))
>          class(obj) <- c(attr(obj, "class"),
>                          if(right) c(right = "noquote") else
"noquote")
>      obj
> }
> Notice what happens with right = TRUE:
>> x <- noquote("a", right = TRUE)
>> x
> [1] a
>> class(x)
>      right
> "noquote"
> The class vector for x is named.  The print method pays attention to the 
> name, so we get different behaviour for a class of "noquote" and
a class
> of c(right = "noquote").
> I had never noticed a named class vector before, and it raised some 
> questions for me:
> - Is this used anywhere else?
Not that I'd be aware of: I think MMae is the expert here.
> - Are names preserved in all the operations normally done on a class 
> vector?  (As far as I can see they are, but maybe I've missed
something.)
> - Is it a good idea to encode a string value worth of information in the 
> name, rather than setting the class to something like
c("noquote",
> "right") instead?
My preference would be to have unnamed class vectors, so that the names
could perhaps eventually be used to store the name of the package which
owns the class.  For noquote, I guess you'd want something like

  c("noquote_right", "noquote")

Best
-k
> Comments would be welcome.
> Duncan Murdoch
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel