Odd behavior of symbol objects with classes/attributes

On 17/06/2013 1:01 PM, Winston Chang wrote:

I've been trying to add classes and attributes to symbol objects
(created with quote()), and the behavior is very strange, as
illustrated in the examples below.

If symbols aren't meant to have classes and attributes attached to
them, then perhaps R should throw errors when you attempt to do it?

I think this is a consequence of another strange property of symbol objects, 
namely that they are not copied on assignment.  This is also true of 
environments, NULL, builtin function references, and some rarely encountered 
types like external pointers and weak references.  You're allowed to assign 
attributes to all of these other than NULL, but you will find strange things 
happen if you do it (as demonstrated in your code below).

I thought I remembered reading a list of these in the documentation 
somewhere, but when I went to look for it just now, I couldn't find it. 
(Maybe I'm remembering a comment in the source.)  It would be a useful 
addition.

Duncan Murdoch

# Using str() strips class from object
x <- quote(foo)
class(x) <- "bar"
str(x)
# Class 'bar'  symbol foo
str(x)
# symbol foo


# Attempting to overwrite doesn't affect class.
# str() still strips class from the object.
x <- quote(foo)
class(x) <- "bar"
x <- quote(foo)
str(x)
# Class 'bar'  symbol foo
str(x)
# symbol foo


# Changing class of one object affects other
x <- quote(foo)
y <- quote(foo)
class(x) <- "bar"
class(y)
# [1] "bar"
str(y)
# Class 'bar'  symbol foo
str(y)
# symbol foo
str(x)
# symbol foo


# Changing attribute of one object affects other
# Unlike with class, str() doesn't cause other attributes to disappear
x <- quote(foo)
y <- quote(foo)
attr(x, "a") <- "bar"
str(y)
# length 1 foo
# - attr(*, "a")= chr "bar"
str(y)
# length 1 foo
# - attr(*, "a")= chr "bar"
str(quote(foo))
# length 1 foo
# - attr(*, "a")= chr "bar"


-Winston