Skip to content

about the subscript assignment porblem

2 messages · kh zu, Thomas Lumley

Original
kh zu
# 
Hi, Dear,

  I have a problem for the new version 2.0 of R
When I put this code in it give me the error:

uis$ivhx3[uis$ivhx>0] <- 1*(uis$ivhx==3)
Error: NAs are not allowed in subscripted assignments

but this would happen in version 1.91
there are missing data in the data set and it is
represented as NA
the data set can be find on line 
at the ucla web
actually this is a example from its web site

the c code is at
https://svn.r-project.org/R/trunk/src/main/subassign.c

:
if(length(y) > 1)
	for(i = 0; i < n; i++)
	    if(INTEGER(indx)[i] == NA_INTEGER)
		error("NAs are not allowed in subscripted
assignments");
 So maybe we need get rid of this conditional code.

  

  Best Wishes
  Ku
Thomas Lumley
#
On Tue, 30 Nov 2004, kh zu wrote:

            

      
      
      
    

        
What would happen?  In version 1.9.0 (I don't have a 1.9.1 lying around) 
you get a warning if any of the subscripts are FALSE
   > x<-1:3
   > y<-7:9
   > x[c(TRUE,NA,FALSE)]<-y
   Warning message:
   number of items to replace is not a multiple of replacement length

and if all the subscripts are TRUE or NA the NAs are treated as FALSE

   > x<-1:3
   > y<-7:9
   > x[c(TRUE,NA,TRUE)]<-y
   > x
   [1] 7 2 9


The change was largely motivated by numerical indices, where the problem 
is worse.   In particular if x is 1,2,3, i is 1, NA, 2  and y is 7,8,9, 
should
    x[i] <- y
set the second element of x to 8 or 9?  The answer was different for 
vectors and matrices in 1.9.1.

It was clearly a deliberate change, not a mistake (it's hard to see how it 
could have happened accidentally, and it was documented), and the previous 
behaviour was wrong.


 	-thomas