R help - May 2010 - Data frames, passing by value, and performance

I understand that everything passed to an R function is passed "by
value".  This would seem to include data frames, which my current
application uses heavily, both for storing program inputs, and holding
intermediate and final results.  In trying to get greater performance
out of my R code, I am wondering if there is any clean way to access
data frames without having them copied all the time.  Or is my only
option to make them global, and write to them using <<-  ?

I have considered using matrices, but I like the self-documenting aspect
of data frame column names.  Input/output to disk is not the issue here,
as that does not take long in my case.  It's just the internal parameter
passing that I'm concerned about.

(I've checked R-FAQ, R-lang and searched the R-help archives, but didn't
see any specific mentions of this.)

Thanks.

Grant Schultz

If you don't modify the data frame in your function it won't
physically make a new copy.

On Mon, May 24, 2010 at 10:29 AM,  <gschultz at scriptpro.com>
wrote:
> I understand that everything passed to an R function is passed "by
> value". ?This would seem to include data frames, which my current
> application uses heavily, both for storing program inputs, and holding
> intermediate and final results. ?In trying to get greater performance
> out of my R code, I am wondering if there is any clean way to access
> data frames without having them copied all the time. ?Or is my only
> option to make them global, and write to them using <<- ??
>
> I have considered using matrices, but I like the self-documenting aspect
> of data frame column names. ?Input/output to disk is not the issue here,
> as that does not take long in my case. ?It's just the internal
parameter
> passing that I'm concerned about.
>
> (I've checked R-FAQ, R-lang and searched the R-help archives, but
didn't
> see any specific mentions of this.)
>
> Thanks.
>
> Grant Schultz
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

R is pretty smart about duplicating only when necessary. That is,
arguments passed to a function are copy-on-write. Also, I think (someone
more knowledgeable please correct if I'm wrong) it may be better to use
the data frame, which is just a list internally, because if you only
modify one column, only that column is duplicated, not the entire data
frame. If you were to use a matrix, the entire matrix would require
duplication.

-Matt

On Mon, 2010-05-24 at 09:29 -0500, gschultz at scriptpro.com
wrote:
> I understand that everything passed to an R function is passed "by
> value".  This would seem to include data frames, which my current
> application uses heavily, both for storing program inputs, and holding
> intermediate and final results.  In trying to get greater performance
> out of my R code, I am wondering if there is any clean way to access
> data frames without having them copied all the time.  Or is my only
> option to make them global, and write to them using <<-  ?
> 
> I have considered using matrices, but I like the self-documenting aspect
> of data frame column names.  Input/output to disk is not the issue here,
> as that does not take long in my case.  It's just the internal
parameter
> passing that I'm concerned about.
> 
> (I've checked R-FAQ, R-lang and searched the R-help archives, but
didn't
> see any specific mentions of this.)
> 
> Thanks.
> 
> Grant Schultz
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.