[R] within: order of newly added variables

[Duncan Murdoch]

On 9/2/2009 9:27 AM, RINNER Heinrich wrote:
> Dear R community,
> 
> I am using function 'within' in R.2.9.1 to add variables to an existing data.frame. This works wonderful, except for one minor point: The new variables are added to the data in reverse order.
> 
> For example:
> x <- data.frame(a = 1:3, b = 4:6)
> y <- within(x, {
>      c = a^2
>      d = b^2
>      e = c+d
>      }
> )
> gives
>   a b  e  d c
> 1 1 4 17 16 1
> 2 2 5 29 25 4
> 3 3 6 45 36 9
> 
> Just what I want; except that I would prefer the columns to be in order a,b,c,d,e instead.
> 
> I could use transform ("transform(x, c=a^2, d=b^2, e=c+d)"), which preserves the specified order of variables, but that won't work here because unfortunately it doesn't find object 'd' (same with "within(x, {e = c+d; d = b^2; c = a^2})", by the way).
> 
> Of course in my toy example I can easily do something like y[, c(1:2,5:3)] afterwards, but I'd like to ask if maybe there is a simple way to make 'within' preserve the order of specified variables (or maybe someone can shed light on why this is not possible?).

Environments have no defined order, so you should not expect the order 
of evaluations which create new variables to produce them in a 
reproducible order.

To get the order you want, declare it from the beginning when you create 
the data.frame.  Or if x came from somewhere else, you could do this by 
adding the new columns first, then modifying them:

x <- data.frame(a = 1:3, b = 4:6)  # original creation
x <- data.frame(x, c=1,d=1,e=1)    # add some new columns
y <- within(x, {                   # modify them
      c = a^2
      d = b^2
      e = c+d
      }
)

Duncan Murdoch