[R] Question about <<- assignment

[Duncan Murdoch]

On 01/07/2009 9:34 PM, Rolf Turner wrote:
> On 2/07/2009, at 12:20 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.
> 
> I would agree that the documentation would *suggest* that, but it
> doesn't explicitly say it, and no more it should, since the "[<<-"
> operator does not exist.  Unlike the "[<-" operator which does exist.
> 
> Moreover there is little reason for a "[<<-" operator to exist.
> 
> The basic raison-d'etre for the "<<-" operator is to allow one to
> assign objects in the global environment from within a function.
> But you shouldn't do it, even though it is possible!!!

No, I'd say that's an unfortunate compromise.  The real use is for 
nested evaluation:  a little worker function defined within a bigger 
function needs to access the parent's variables for reading (no problem, 
lexical scope gives that) and writing (that's what <<- is for).

So what to do when the target of <<- doesn't exist?  I'd say a better 
solution than we have is for it to fail, rather than assign in the 
global environment.

Duncan Murdoch

> 
> Modifying a component of an existing object in the global environment
> from within a function is even more to be avoided, it seems to me.  And
> the designers of R seem to have drawn the line at this point.
> 
> There may be conceptual difficulties with implementing a "[<<-"  
> operator,
> over and above the philosophical position that using such an operator
> would be bad practice.  I don't know.  But I doubt that R Core will move
> to build a "[<<-" operator, even should it prove to be possible.
> 
> The error message ``object "z" not found'' is certainly misleading.   
> I don't
> know how hard it would be to change the code so that a more enlightening
> error message gets issued.  Probably not worth the effort since few  
> users
> would ever entertain the idea of using "[<<-".
> 
> 	cheers,
> 
> 		Rolf Turner
> 
> ######################################################################
> Attention:\ This e-mail message is privileged and confid...{{dropped:9}}
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.