R help - Mar 2012 - index values of one matrix to another of a different size

> Hello,
>
> Is this the fastest way to use indices from one matrix to reference rows
> in another smaller matrix? I am dealing with very big data (lots of columns
> and I have to do this lots of times).
>
> ######sample data ##############
> vals = matrix(LETTERS[1:9], nrow=3,ncol=3)
> colnames(vals) = c('col1','col2','col3')
> rownames(vals) = c('row1','row2','row3')
> > vals
> col1 col2 col3
> row1 "A"  "D"  "G"
> row2 "B"  "E"  "H"
> row3 "C"  "F"  "I"
>
> # this is a matrix of row references to vals above. The values all stay in
> the same column but shift in row via the indices.
>  indx = matrix(c(1,1,3,3,2,2,2,3,1,2,2,1),nrow=4,ncol=3)
> > indx
> [,1] [,2] [,3]
> [1,]    1    2    1
> [2,]    1    2    2
> [3,]    3    2    2
> [4,]    3    3    1
> ############### end sample data ####################
>
> # my solution
> >
>
matrix(vals[cbind(c(indx),rep(1:ncol(indx),each=nrow(indx)))],nrow=nrow(indx),ncol=ncol(indx))
> [,1] [,2] [,3]
> [1,] "A"  "E"  "G"
> [2,] "A"  "E"  "H"
> [3,] "C"  "E"  "H"
> [4,] "C"  "F"  "G"
>
> Thanks,
>
> Ben
>
> PS - Rui - I thought you may want to see this since I think this will be a
> faster way to deal with the issue you were working with me on...although I
> don't show how I build the matrix of indices, I think you get the idea.
>
>
	[[alternative HTML version deleted]]

Hello,

I don't know if it's the fastest but it's more natural to have an
index
matrix with two columns only,
one for each coordinate. And it's fast.

fun <- function(valdata, inxdata){
	nr <- nrow(inxdata)
	nc <- ncol(inxdata)
	mat <- matrix(NA, nrow=nr*nc, ncol=2)
	i1 <- 1
	i2 <- nr
	for(j in 1:nc){
		mat[i1:i2, 1] <- inxdata[, j]
		mat[i1:i2, 2] <- rep(j, nr)
		i1 <- i1 + nr
		i2 <- i2 + nr
	}
	matrix(valdata[mat], ncol=nc)
}

fun(vals, indx)

Rui Barradas


--
View this message in context:
http://r.789695.n4.nabble.com/Re-index-values-of-one-matrix-to-another-of-a-different-size-tp4458666p4460575.html
Sent from the R help mailing list archive at Nabble.com.