Comments on: Design Flaws in R #2 — Dropped Dimensions

By: Anonymous

Anonymous — Thu, 02 May 2013 20:17:06 +0000

Thank you, this is a very useful tip! I have an automated porgram that has to deal with this situation often. The “drop” statement shortened my program by at least 20 lines!

By: Clement Kent

Clement Kent — Wed, 06 Mar 2013 23:06:17 +0000

My previous comment on APL, the predecessor of R and S, applies well to Eli’s complaint. APL, and a number of non-S languages that descend from it, do not drop dimensions, have rational designs for permutation, transposition, and subsetting, and allow a larger number of operations along dimensions of arrays than R. I’m not going backward to use APL, but really wish R could be upgraded to be as good as its grandparent is for arrays.

By: Radford Neal

Radford Neal — Wed, 06 Mar 2013 20:29:41 +0000

For the three-dimensional case, you maybe should use the “aperm” function.

By: Eli

Eli — Wed, 06 Mar 2013 20:13:34 +0000

I too write a lot of programs in R that use matrix multiplication.

#this is an example that kills me
a=matrix(1,4,4)
# transpose of the first row is a 4×1 COLUMN matrix
t(a[1,])

sadly, first R removes the dim attr, then does a transpose on the vector (which by some bizarre logic is possible with no dim attr) and gives you a 1×4 ROW matrix.

Horrible. However this is easily circumvented by employing drop=FALSE defensively throughout all your R code. Until… you need to work with arrays with more than 2 dimensions.

a=array(1,dim=c(1,2,10))
#once again R drops the dim attr and we get a row
t(a[,,1])
#that’s ok, let’s use drop=FALSE
t(a[,,1,drop=FALSE])
#oh no, a is not a matrix so we get an error
#–which actually makes sense unlike the behavior for t() on a vector with no dim attr

But what do we do to get a[,,1] as a matrix so we can take the transpose?

#this works but is computationally 3x as expensive as needed
b=a[,,1]
dim(b)=c(dim(a)[1],dim(a)[2])
t(b)

#this is less expensive but ugly
matrix(a[,,1],dim(a)[2],dim(a)[1],byrow=TRUE)

#by the way, this does not work for getting the correct transpose of a matrix
b=a[,,1]
dim(b)=c(dim(a)[2],dim(a)[1])

My solution was to write my own utility function for subsetting 3D matrices.

By: Radford Neal

Radford Neal — Wed, 24 Oct 2012 22:24:22 +0000

Well, S and APL both have vectors, but beyond that, any connection is VERY loose…

By: Clement Kent

Clement Kent — Wed, 24 Oct 2012 21:41:30 +0000

A bit of history: S was loosely based on the much more rigorous language APL. APL is more consistent about dimensions than either S or R and is/was, IMHO, easier to use for most array manipulations. On the other hand, R is free and comes with a gazillion great packages; APL is now used by a much smaller group of programmers than R.

By: Radford Neal

Radford Neal — Tue, 06 Dec 2011 19:41:49 +0000

Yes, but matrix(M[1:i,],ncol=ncol(M),byrow=TRUE) is even worse than M[1:i,,drop=FALSE].

By: denes turei

denes turei — Tue, 06 Dec 2011 01:23:07 +0000

or you can use a function
as.matrix()
or
matrix(vector,ncol=a,byrow=T)

By: Florian B

Florian B — Thu, 12 Aug 2010 16:12:47 +0000

Regarding the disambiguation of vectors versus matrixes:

A matrix is just a vector with a dim attribute:
> v v
[1] 1 2 3 4 5 6 7 8 9 10
> attr(v,”dim”) v
[,1] [,2] [,3] [,4] [,5]
[1,] 1 3 5 7 9
[2,] 2 4 6 8 10

There are loads of things in R which are confusing in the beginning, but once you know them, often you can work with them.

By: Radford Neal

Radford Neal — Fri, 13 Nov 2009 00:31:18 +0000

I think “data analysis” versus “programming” is a false distinction. A data analysis environment that doesn’t allow programming would be very limiting. That’s one reason why R is popular – you can use it for data analysis, which includes a bit of programming now and then when you need to (for example) do some complicated conversion from one data format to another.

I also think that there is little or no actual conflict between good features for data analysis and good features for programming. Historically, S was designed first with data analysis in mind, with probably little thought for programming. And I’m not blaming the original designers for not forseeing how programming would become more important. But if S/R were re-designed from scratch, there would be no reason to put in features bad for programming in order to make data analysis easier. It’s not too hard to make things good for both. Even now, it may not be hard, as seen from my other posts on how to solve these problems.