[R] Question about <<- assignment

[Duncan Murdoch]

On 01/07/2009 8:20 PM, Hsiu-Khuern Tang wrote:
> Is this expected behavior?
> 
>> z <- 1:5
>> z[1] <<- 0
> Error in z[1] <<- 0 : object "z" not found
> 
> The documentation seems to suggest that z will be found in the global
> environment and modified accordingly.

z <<- 0 works as documented, it's the indexing that fails.  (I tried 
other replacement functions and they also failed.)  I think this case is 
also behaving as documented, but the docs are in the R Language 
Definition manual.  To evaluate

z[1] <<- 0

the following sequence is what is supposed to happen:

   - Look up z in the parent environment of the evaluation frame, and 
assign to a temporary variable.  (This is what fails; z is in the 
evaluation frame, not its parent.)
   - Set the first element of the temporary.
   - Use <<- to assign the temporary to z.

This might be surprising, but I think it makes sense. Using <<- when 
evaluating in the global environment is almost certainly an error; 
shouldn't R signal it as such?

Duncan Murdoch