Skip to content

Iteration idioms & laziness

3 messages · Alexander Søndergaard, Uwe Ligges, Duncan Murdoch

Alexander Søndergaard
# 
Hi all,

I'm new to R. Having a functional background, I was wondering what's
the idiomatic way to iterate. It seems that for loops are the default
given there's no tail-call optimization.

I'm curious to know whether there is a way to transform the following
toy snippet into something that doesn't eat up gigabytes of memory
(like it's for loop counterpart) using laziness:

Reduce('+', seq(1,1e6))

Thanks!

Best regards,
A.S.
1 day later
Uwe Ligges
#
Alexander S?ndergaard wrote:
In this case

   sum(as.numeric(seq(1, 1e6)))

or according to Gauss

   1e6 * (1e6+1) / 2

but for more complicated problems it might be sensible to go the loop 
way ...

Uwe Ligges
Duncan Murdoch
#
On 27/11/2009 3:36 PM, Alexander S?ndergaard wrote:
I believe the iterators and foreach packages give ways to iterate 
without creating the whole array, so they might do what you want.  But 
is the allocation really a problem on modern computers?  The for loop 
version of your example, i.e.

total <- 0
for (i in seq(1,1e6)) total <- total + i
total

uses about 4 megabytes of memory, not "gigabytes".  If you increase the 
limit from 1e6 to 1e9 you'll get gigabytes, but it'll probably take tens 
of minutes to finish running on a system that can run it:  the big 
problem is the interpreted looping, not the memory use.

R has always followed the strategy of making it easy to jump out to C 
code when speed matters, so the natural idiom for your problem is to use 
sum(), not to program it yourself.  In more complicated examples, or if 
you really do want to sum the integers from 1 to a billion on a system 
with limited memory, you'll have to write the C yourself, but it's not hard.

Duncan Murdoch