[R-SIG-Mac] A shorthand for '<-'

[John Chambers]

Duncan Murdoch wrote:
> On 11/21/2007 1:19 PM, Simon Urbanek wrote:
>> On Nov 21, 2007, at 11:45 AM, John Chambers wrote:
>>
>>> Simon Urbanek wrote: James,
>>>
>>> On Nov 21, 2007, at 8:22 AM, James Milks wrote:
>>>
>>>> The '=' sign can be used in place of '<-'.  That's the only  
>>>> shorthand I know for R.
>>>>
>>>
>>> That's not a shorthand. `=` and `<-` are semantically different in R.
>>> Well, not really :-)
>>>
>>> It's true that the "=" operator won't be displayed as "<-", which I  
>>> agree was the original point.  But both assignment operators map  
>>> into the same internal C code, if you di g into the implementation.
>>>
>>
>> I think we may be talking about different things here.
>>
>>  > a=list(a=1,b=2)
>>  > ls()
>> [1] "a"
>>
>>  > a<-list(a<-1,b<-2)
>>  > ls()
>> [1] "a" "b"
>>
>> So `<-` and `=` are *not* semantically equivalent (where `<-` and 
>> `=`  represent symbols in the parse tree).
Ouch!  That's a good example of why people complained when "=" was added 
as an operator.  In your first expression, "=" is being used in two 
quite different ways.  (You can make out a case that argument 
specification is a little like assignment, since it creates an object of 
that name in the environment of the call, but that's pretty subtle.)

In a sense "=" inside a function call is not semantically an operator at 
all; it's absorbed by the parser into the internal structure that 
represents the call.  In contrast, putting an assignment operator inside 
an argument to a function is very different (and, IMO, something to be 
avoided in writing clear S-language code.)

The difference can be highlighted by looking at the parsed but 
unevaluated expressions in the two cases:

 > e1 = quote(list(a=1,b=2))
 > e1[[2]]
[1] 1
 > names(e1)
[1] ""  "a" "b"
 > e2 = quote(list(a<-1, b<-2))
 > e2[[2]]
a <- 1

The "=" has in a sense disappeared, only affecting the names attribute 
of the language object.

>>
>> And I suppose the misunderstanding comes from the interpretation of  
>> `=` and `<-`: I meant them as symbols (which is what I would expect  
>> since we're talking about writing R code) and you interpreted them 
>> as  operators (which cold be expected given that I used backticks 
>> which  was not wise ;)). You are right that `=` and `<-` are 
>> equivalent as  operators:
>>
>>  > `=`(a,list(`=`(a,1),`=`(b,2)))
>>  > ls()
>> [1] "a" "b"
>>
>> I hope this makes things even more clear ;).
>
> Of course, even using them as operators the parser doesn't think they 
> are identical:
>
> > if (a <- 1) cat("yes\n")
> yes
> > if (a = 1) cat("yes\n")
> Error: unexpected '=' in "if (a ="
> > if (`=`(a,1)) cat("yes\n")
> yes
Again, the parser prohibition comes from catching a classic programming 
error in C (and another complaint about using "=" as an assignment 
operator).

while(flag = 0) { // of course, I meant (flag == 0)
  ...
  if(something(...)) flag = 1;
}

after which the programmer wonders why the loop never exits.

(These points recall some surprisingly religious arguments 10 years or 
so ago ;-))
>
> Duncan Murdoch
>