gistfile1.txt

My message:
This was tweeted recently by an R user, and after trying to figure out why this is, I am stumped:
> a <- 5; b <- 1; c <- 4
> f <- function (n) for (i in 1:n) d <- 1/{a*{b+c}}
> g <- function (n) for (i in 1:n) d <- 1/(a*(b+c))
> system.time(f(1000000))
 user  system elapsed
 3.92    0.00    3.94
> system.time(g(1000000))
 user  system elapsed
 4.17    0.00    4.17
(from http://radfordneal.wordpress.com/2010/08/15/two-surpising-things-about-r/)

Anyone know why curly braces are evaluated faster? 




Duncan's response (he comments that "there is more to investigate as to the particular paths through the evaluator"): 

I'll bite and give a slightly informal description of what
I see as going on.

( and { are quite different elements of the language.
For instance,
 (1+2; 3)
is an error, but

 {1+2; 3}

evaluates to 3 and evaluates both "terms" within the expression.

As a result, the evaluation path within the interpreter is quite different.
Specifically, the 1/(2+3) ends up with 3 calls to the C routine Rf_evalList()
with terms (2+3) and (2, 3).

1/{2+3}  ends up with only two calls Rf_evalList(), evaluating with

 (1, {4 + 5})
and
  (4 , 5)

So there is a different and slightly shorter path, but small overhead
per call.

(BTW, I found this via gdb, the debugger for compiled code.)

 D.