R devel - Jul 2011 - Best practices for writing R functions

I am developing an R package for internal use, and eventually for public
release. My understanding is that there is no easy way to avoid copying
function arguments in R (i.e. we don't have the concept of pointers in R),
which makes me wary of freely creating chains of function calls since each
function call implies data copy overhead.

Is the above assessment fair? Are there any good write-ups on best practices
for writing efficient R libraries that take into consideration the
above-mentioned limitations, and any others that might exist?

Thank you,
Alireza


--
View this message in context:
http://r.789695.n4.nabble.com/Best-practices-for-writing-R-functions-tp3686674p3686674.html
Sent from the R devel mailing list archive at Nabble.com.

From my personal experience and following this list some for a 
few years, the best practice is initially to ignore the compute time 
question, because the cost of your time getting it to do what you want 
is far greater, at least initially.  Don't worry about compute time 
until it becomes an issue.  When it does, the standard advice I've seen 
on this list is to experiment with different ways of writing the same 
thing in R, guided by "profiling R code", as described in the
"Writing R
Extensions" manual.  (Googling for "profiling R code" identified
examples.)


       Hope this helps.
       Spencer Graves


On 7/22/2011 6:26 AM, Alireza Mahani wrote:
> I am developing an R package for internal use, and eventually for public
> release. My understanding is that there is no easy way to avoid copying
> function arguments in R (i.e. we don't have the concept of pointers in
R),
> which makes me wary of freely creating chains of function calls since each
> function call implies data copy overhead.
>
> Is the above assessment fair? Are there any good write-ups on best
practices
> for writing efficient R libraries that take into consideration the
> above-mentioned limitations, and any others that might exist?
>
> Thank you,
> Alireza
>
>
> --
> View this message in context:
http://r.789695.n4.nabble.com/Best-practices-for-writing-R-functions-tp3686674p3686674.html
> Sent from the R devel mailing list archive at Nabble.com.
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

On Jul 22, 2011, at 9:26 AM, Alireza Mahani wrote:
> I am developing an R package for internal use, and eventually for public
> release. My understanding is that there is no easy way to avoid copying
> function arguments in R (i.e. we don't have the concept of pointers in
R),
> which makes me wary of freely creating chains of function calls since each
> function call implies data copy overhead.
> 
> Is the above assessment fair?
No. Although R maintains the illusion of pass-by-value, it does to considerable
length to prevent copying if not needed:
> r=rnorm(1e6)
> tracemem(r)
[1] "<0x103fa2000>"
> f = function(x) x
> invisible(f(f(f(f(f(f(r)))))))
As you see not a single copy is created even though there are 6 function calls.

As Spencer said, you should not worry unless you see a sign of a problem.

Cheers,
Simon

> Are there any good write-ups on best practices
> for writing efficient R libraries that take into consideration the
> above-mentioned limitations, and any others that might exist?
> 
> Thank you,
> Alireza
> 
> 
> --
> View this message in context:
http://r.789695.n4.nabble.com/Best-practices-for-writing-R-functions-tp3686674p3686674.html
> Sent from the R devel mailing list archive at Nabble.com.
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
> 
>