Skip to content

Floating point excepting when cbind()ing a matrix of grobs (or environments) with a 0-column matrix

3 messages · Hadley Wickham, William Dunlap, Brian Ripley

Original
Hadley Wickham
# 
library(grid)
e <- rectGrob()
# OR:
# e <- environment()

a <- matrix(list(e), ncol = 1, nrow = 2)
b <- matrix(ncol = 0, nrow = 2)

cbind(a, b)

cbind(a, b)

This reliably crashes R for me.

I realise this is a rather esoteric error condition, but it crops up
for me when creating matrices of grobs to be turned into a ggplot2
plot.

Hadley
William Dunlap
# 
I think the following patch to src/main/bind.c fixes it up.
If the length of an argument to [cr]bind is 0 then it wont't
try to add any data from it to the output any more, just as
is done with the other modes of data.  It was getting the FPE
from i%k when k==0.

Index: src/main/bind.c
===================================================================
--- src/main/bind.c     (revision 47617)
+++ src/main/bind.c     (working copy)
@@ -1218,10 +1218,12 @@
                case LISTSXP:
                    PROTECT(u = coerceVector(u, mode));
                    k = LENGTH(u);
-                   idx = (!isMatrix(u)) ? rows : k;
-                   for (i = 0; i < idx; i++)
-                       SET_VECTOR_ELT(result, n++,
-                                      duplicate(VECTOR_ELT(u, i % k)));
+                   if (k>0) {
+                       idx = (!isMatrix(u)) ? rows : k;
+                       for (i = 0; i < idx; i++)
+                           SET_VECTOR_ELT(result, n++,
+                                          duplicate(VECTOR_ELT(u, i %
k)));
+                   }
                    UNPROTECT(1);
                    break;
                default:

Bill Dunlap
TIBCO Software Inc - Spotfire Division
wdunlap tibco.com
1 day later
Brian Ripley
# 
Thank you: I've applied something similar (this was another in-line 
patch that did not apply, even when I fixed the line wrap).
On Mon, 19 Jan 2009, William Dunlap wrote: