Skip to content

empty array

6 messages · Christophe Genolini, Gábor Csárdi, Brian Ripley +2 more

Christophe Genolini
# 
Hi the list

Is it possible to create an empty matrix ? I do not mean an matrix with 
a single value that is NA (which is not empty) but a real empty one, 
with length=0.

I do not understand why we have length(numeric()), length(factor()) and 
length(character()) to zero, and length(array()) to one... Any rason for 
that ?

Thanks

Christophe
Gábor Csárdi
#
On Sat, Mar 15, 2008 at 04:33:32PM +0100, Christophe Genolini wrote:
Sure:
[,1] [,2] [,3] [,4] [,5]
[1] 5 4 0
[1] 0
That i don't know, maybe someone else does.

G.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Sat, 15 Mar 2008, Gabor Csardi wrote:

        

            

      
      
      
    

        
See the reference on the help page for array and section 3.3.3 of the R 
FAQ (about the 16th item).  When R was first developed, S did not allow 
zero dimensions and so the minimum length for an array (or matrix) was 
one.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      John Kane
    

    

      
      #
    

  


  

      
Will
mymatrix <- NULL do what you want?

        
--- Christophe Genolini <cgenolin at u-paris10.fr> wrote:

        

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Christophe Genolini
    

    

      
      #
    

  


  

      
John Kane a ?crit :

            

      
      
      
    

        
Well, in your code, 'mymatrix' is not a matrix :

a<-array(dim=c(0,0))  # Solution of Gabor Csardi
<0 x 0 matrix>
class(a)
[1] "matrix"
b<-NULL
class(b)
[1] "NULL"

Your definition will probably works in most case, but in S4, the uses of 
type is an important point, so I need to use matrix where I shall use 
matrix, even if it is an empty one.

Christophe

            

      
      
      
    

        

      
      
    

            

      
      
      
    

  
  


  


      

    

    
2 days later

    

      
      

        


  

        

      robin hankin
    

    

      
      #
    

  


  

      
Hello everyone

I know other, more knowledgeable, people
have replied to Christophe's question, but
perhaps the List would be interested to know
that zero-extent arrays are useful (to me at least)
  because although such an array  has no content, the
dimname are nevertheless retained:


 > a <- array(0,dim=c(0,3))
 > dimnames(a) <- list(size=c() , fish=c("cod","skate","crab"))
 > b <- array(0,dim=c(2,0))
 > dimnames(b) <- list(size=c("huge","small"),depth=c())
 >


We can attach these arrays ---both of  which are of length 0---using  
adiag():

 > library(magic)
 > adiag(a,b)
        fish
size    cod skate crab
   huge    0     0    0
   small   0     0    0
 >


Note how the dimnames of "a" and "b" are retained in the output.
The contents of this array are the default "pad" value of adiag().

This is terribly useful in the humble workaday world
of high-dimensional magic hypercubes.




rksh

        
On 15 Mar 2008, at 15:33, Christophe Genolini wrote:

        

            

      
      
      
    

        

      
      
    

        
--
Robin Hankin
Uncertainty Analyst and Neutral Theorist,
National Oceanography Centre, Southampton
European Way, Southampton SO14 3ZH, UK
  tel  023-8059-7743