[R] Augment 'Matrix' matrices

[Martin Maechler]

>>>>> "SH" == Scott Hyde <hydes at byuh.edu>
>>>>>     on Mon, 18 Jun 2007 16:59:00 -1000 (HST) writes:

    SH> Martin, How does Matrix implement augmented matrices?  I
    SH> tried this and got the expected result:


{Replying to  R-help,  since this question has come up several
 times }

    >> V=matrix(1,2,3)
    >> V=cbind(V,V)
    >> V
    SH>      [,1] [,2] [,3] [,4] [,5] [,6]
    SH> [1,]    1    1    1    1    1    1
    SH> [2,]    1    1    1    1    1    1

    SH> But when I did it with Matrix instead I got:

    >> library(Matrix)

    >> V=Matrix(1,2,3)
    >> V=cbind(V,V)
    >> V
    SH> V V
    SH> [1,] ? ?

cbind() and rbind() cannot work with S4 objects because their
first formal argument is  '...'
[ and with S3 objects they only work insofar as S3 generics can
  "work": i.e. they only "work" when the first argument is of the
  respective class, but fail, e.g. for  cbind(1, <object>)
  when <object> is of a non-standard S3 class.
]
In earlier versions of Matrix, there was a sophisticated "hack"
that made  cbind() and rbind()   "work".

But because it was a hack, and some people called it "horrible"
rather than "sophisticated", we had to give it up.
{well, the really compelling argument was an example of
 do.call(rbind, <list of length 1000>) which was *very* inefficient}

Instead, cbind2() and rbind2() have been written a
few R versions ago to be used as (S4) generic functions. 
-->  help(cbind2)

In 'Matrix', we also define cBind() and rBind()  to be used as
direct (n-argument) substitutes for cbind() or rbind(),
respectively.

Martin