R help - Jan 2009 - Memory allocation

Hello everyone,

I have the following issue: one function generates a very big array (can be
more than 1 Gb) and returns a few variables, including this big one. Memory
allocation is OK while the function is running, but the final steps make
some copies that can be problematic. I looked for a way to return the values
without copying (even tried Rmemprof), but without success. Any ideas?
The code looks like this:

myfunc <- function() {
...
bigarray <- ...
...
final <- list(..., bigarray=bigarray, ...)
class(final) <- "myfunc"
final
}

Thank you in advance,
Gabriel.

	[[alternative HTML version deleted]]

On 1/16/2009 12:46 PM, Gabriel Margarido wrote:
> Hello everyone,
> 
> I have the following issue: one function generates a very big array (can be
> more than 1 Gb) and returns a few variables, including this big one. Memory
> allocation is OK while the function is running, but the final steps make
> some copies that can be problematic. I looked for a way to return the
values
> without copying (even tried Rmemprof), but without success. Any ideas?
> The code looks like this:
> 
> myfunc <- function() {
> ...
> bigarray <- ...
> ...
> final <- list(..., bigarray=bigarray, ...)
> class(final) <- "myfunc"
> final
> }
> 
> Thank you in advance,
I believe this will do less copying, but I haven't profiled it to be 
sure.  Replace the last three lines with this one statement:

structure(list(..., bigarray=bigarray, ...),
                    class = "myfunc")

If that doesn't help, then you really need to determine where the 
copying is happening: you can use Rprofmem() to do that.


Duncan Murdoch

Gabriel Margarido <gramarga <at> gmail.com> writes:
> ... I looked for a way to return the values
> without copying (even tried Rmemprof), but without success. Any ideas?
> ...
I solved similar problems using the R.oo package, which emulates
pass-by-reference semantics in 'R'.

HTH

Keith