Hi, there is a bug in the calculation of row names in rbind.data.frame. When one of the arguments has 0 rows but is named in the call, this mistakenly contributes an element in the "row.names" attribute of the result, e.g.:
foo <- data.frame(x = 1:10, y = rnorm(10)) bar1 <- rbind.data.frame(foo[1:5,], foo[numeric(0),]) dim(bar1)
[1] 5 2
bar2 <- rbind.data.frame(a = foo[1:5,], b = foo[numeric(0),]) dim(bar2)
[1] 6 2
sessionInfo()
Version 2.3.0 Under development (unstable) (2006-01-17 r37109) i686-pc-linux-gnu attached base packages: [1] "methods" "stats" "graphics" "grDevices" "utils" "datasets" [7] "base" This happens because
rownames(bar1)
[1] "1" "2" "3" "4" "5"
rownames(bar2)
[1] "a.1" "a.2" "a.3" "a.4" "a.5" "b"
I think the following patch (to the 'Make.row.names' function defined
inside rbind.data.frame) fixes this:
-Deepayan
Index: src/library/base/R/dataframe.R
===================================================================
--- src/library/base/R/dataframe.R (revision 37109)
+++ src/library/base/R/dataframe.R (working copy)
@@ -886,7 +886,7 @@
if(nchar(nmi) > 0) {
if(ni > 1)
paste(nmi, ri, sep = ".")
- else nmi
+ else nmi[ri]
}
else if(nrow > 0 && identical(ri, 1:ni))
seq(from = nrow + 1, length = ni)