R help - Apr 2008 - preallocating matrices and rda read-back objects

I've read in Phil Spector's new book that it's a good idea to  
preallocate a big matrix, like

u <- matrix(0,nrow,ncol) # (1)

Now, I read contents of a huge matrix from a Fortran binary dump.

u <- readBin(con,what="double",n=nrow*ncol) # (2)

If I do (1) and then (2), u is a vector, obviously it's either  
reallocated or its matrix nature is lost -- overridden?  overwritten?

Instead, I do it now as

u <-  
matrix(readBin(con,what="double",n=nrow*ncol),nrow=nrow,ncol=ncol) #
(3)

What's going on with memory management here and what's the right way  
to make it efficient -- and how to preallocate?

After that, I'm saving u as R binary object in an rda file.  Does it  
make sense to preallocate u before reading it back now from the rda  
file?

Cheers,
Alexy

Alexy Khrabrov wrote:
> I've read in Phil Spector's new book that it's a good idea to  
> preallocate a big matrix, like
> 
> u <- matrix(0,nrow,ncol) # (1)
> 
> Now, I read contents of a huge matrix from a Fortran binary dump.
> 
> u <- readBin(con,what="double",n=nrow*ncol) # (2)
> 
> If I do (1) and then (2), u is a vector, obviously it's either  
> reallocated or its matrix nature is lost -- overridden?  overwritten?
> 
> Instead, I do it now as
> 
> u <-  
>
matrix(readBin(con,what="double",n=nrow*ncol),nrow=nrow,ncol=ncol) #
(3)
> 
> What's going on with memory management here and what's the right
way
> to make it efficient -- and how to preallocate?
> 
> After that, I'm saving u as R binary object in an rda file.  Does it  
> make sense to preallocate u before reading it back now from the rda  
> file?
>
This kind of preallocation only makes sense for loops. In your case, it 
does not yield benefits.

Anyway, if you really want to reassign things in u, you could say:

u[] <- readBin(con,what="double",n=nrow*ncol) # (2)

(empty brackets for indexing all elements)

Best wishes,
Uwe



> Cheers,
> Alexy
> 
> ______________________________________________
> 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.