R devel - Mar 2004 - cbind/rbind fail on matrixes containing lists (PR#6702)

Today's R 1.9.0 beta:
> m1
     [,1] [,2] [,3] [,4]
[1,] NA   NA   NA   NA  
[2,] NA   NA   NA   NA  
[3,] NA   NA   NA   NA  
[4,] NA   NA   NA   NA  
> class(m1[1,1])
[1] "list"
> cbind(m1,m1)
Error in cbind(...) : cannot create a matrix from these
types
> rbind(m1,m1)
Error in rbind(...) : cannot create a matrix from these types
> version
         _                   
platform sparc-sun-solaris2.8
arch     sparc               
os       solaris2.8          
system   sparc, solaris2.8   
status   beta                
major    1                   
minor    9.0                 
year     2004                
month    03                  
day      26                  
language R                   


Gregory R. Warnes
Manager, Non-Clinical Statistics
Pfizer Global Research and Development
Tel: 860-715-3536



LEGAL NOTICE\ Unless expressly stated otherwise, this messag...{{dropped}}

Yep, bind.c:do_bind has

    switch(mode) {
    case NILSXP:
    case LGLSXP:
    case INTSXP:
    case REALSXP:
    case CPLXSXP:
    case STRSXP:
        break;
    default:
        errorcall(call, "cannot create a matrix from these types");
    }


Looks like a fairly simple fix to me.

	-thomas

There are no `matrixes containing lists'.  There are lists which are also 
matrices, but their contents are the vector elements  (unless you meant 
that it was a list of lists, which seems not be the case in your example).
In the absence of a reproducible example, it is not clear to me exactly 
what you want to do.

Here is a list matrix example:

A <- matrix(as.list(1:4), 2, 2)
cbind(A, A)

It appears to be intentional (line 950 of bind.c and the actual cbind
code), in which case this is a documentation bug and I have altered the
docs.


On Fri, 26 Mar 2004 gregory_r_warnes@groton.pfizer.com wrote:
> 
> Today's R 1.9.0 beta:
> 
> > m1
>      [,1] [,2] [,3] [,4]
> [1,] NA   NA   NA   NA  
> [2,] NA   NA   NA   NA  
> [3,] NA   NA   NA   NA  
> [4,] NA   NA   NA   NA  
> > class(m1[1,1])
> [1] "list"
Why not class(m1)?  It should be the same ....
> > cbind(m1,m1)
> Error in cbind(...) : cannot create a matrix from these types
> > rbind(m1,m1)
> Error in rbind(...) : cannot create a matrix from these types
> 
> > version
>          _                   
> platform sparc-sun-solaris2.8
> arch     sparc               
> os       solaris2.8          
> system   sparc, solaris2.8   
> status   beta                
> major    1                   
> minor    9.0                 
> year     2004                
> month    03                  
> day      26                  
> language R                   

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

On Fri, 26 Mar 2004, Thomas Lumley wrote:
> On Fri, 26 Mar 2004, Thomas Lumley wrote:
> 
> >
> > Yep, bind.c:do_bind has
> >
> >     switch(mode) {
> >     case NILSXP:
> >     case LGLSXP:
> >     case INTSXP:
> >     case REALSXP:
> >     case CPLXSXP:
> >     case STRSXP:
> >         break;
> >     default:
> >         errorcall(call, "cannot create a matrix from these
types");
> >     }
> >
> >
> > Looks like a fairly simple fix to me.
> 
> Not any more it doesn't.  bind.c:AnswerType puts lists together with
> pairlists and who knows what else in its `other' class, so a bit more
> thought and coding is needed.
> 
> It should be fairly straightforward, but it's not a localised change.
That's not the main problem area: the actual cbind code does not handle
vector lists.  I've treated it as a documentation fix / enhancement 
request.

Brian

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Hello Brian,
> -----Original Message-----
> From: Prof Brian Ripley [mailto:ripley@stats.ox.ac.uk]
> 
> There are no `matrixes containing lists'.  There are lists 
> which are also 
> matrices, but their contents are the vector elements  (unless 
> you meant 
> that it was a list of lists, which seems not be the case in 
> your example).
> In the absence of a reproducible example, it is not clear to 
> me exactly 
> what you want to do.
> 
> Here is a list matrix example:
> 
> A <- matrix(as.list(1:4), 2, 2)
> cbind(A, A)
> 
I was, in fact, using a matrix containing "list" elements.  

I did fail to provide the definition of my 'list matrix'.  It was
simply:
> m1 <- matrix(list(NA),4,4)
> m1
     [,1] [,2] [,3] [,4]
[1,] NA   NA   NA   NA  
[2,] NA   NA   NA   NA  
[3,] NA   NA   NA   NA  
[4,] NA   NA   NA   NA  
> class(m1)
[1] "matrix"
> class(m1[1,1])
[1] "list"
> cbind(m1,m1)
Error in cbind(...) : cannot create a matrix from these
types
> rbind(m1,m1)
Error in rbind(...) : cannot create a matrix from these types
> It appears to be intentional (line 950 of bind.c and the actual cbind
> code), in which case this is a documentation bug and I have 
> altered the
> docs.
Odd.
> Why not class(m1)?  It should be the same ....
But it is not:
> class(m1)
[1] "matrix"
> class(m1[1,1])
[1] "list"

-Greg


LEGAL NOTICE\ Unless expressly stated otherwise, this messag...{{dropped}}