January 2010 Archives

I've been working on a number of projects recently in R and find it fairly frustrating. There's no way to modify a variable in a function (as in the case of languages that support closures). This makes it especially hard for a language like R that has pretty good functional programming support.

I'm currently working on a library to make R act more sanely (called SaneR - I'll post again when I make it available online).  One function that I've found quite useful lately is one I call "closure":

closure <- function(..., frame=3) {
  parent <- parent.frame()
  frame <- ifelse(frame=="sapply", 4, frame)
  grandparent <- sys.frame(sys.nframe()-frame)
  for(name in c(...))
    assign(name, get(name, envir=parent), envir=grandparent)
}

Using this function you can affect variables outside of a function's scope.  For instance:

x <- 0
lapply(list(1,2,3,4,5), function(y) {
   x <- x + 1
   message("Run number ", x)
   closure("x")
   print(y);
})
print(x)

By calling closure() the variable x is set in the parent (actually, grandparent's) environment, meaning that by the time you reach the print statement x is 5.  This is great for things like counters within lapply. 

Other things that are going into the SaneR packages are:

• a doTimes like Ruby's <int>.times { } method that does something some number of times and returns the results (like calling a map function to an integer range)
• A foreach that can handle lists sanely.  For instance, if you want to do something for every name and value in a list, it takes more lines of code then it should.
• A propper logging mechanism for long running tasks.  A whole lot of R tasks take a while to run (R is slow).  I'm writing a logging class that can print out the % complete for some task as well as an estimate to time of completion.
• and more....

If there are any more ideas out there, comment on this post.