rbind data frames, duplicated rownames issue

Question

While duplicated row (and column) names are allowed in a matrix, they are not allowed in a data.frame. Trying to rbind() some data frames having row names in common highlights this problem. Consider two data frames below:

foo = data.frame(a=1:3, b=5:7)
rownames(foo)=c("w","x","y")
bar = data.frame(a=c(2,4), b=c(6,8))
rownames(bar)=c("x","z")
# foo               bar
#   a b               a b
# w 1 5             x 2 6
# x 2 6             y 4 8
# y 3 7

Now trying to rbind() them (Pay attention to the row names):

rbind(foo, bar)
#    a b
# w  1 5
# x  2 6
# y  3 7
# x1 2 6
# z  4 8

But for the case of matrix:

rbind(as.matrix(foo), as.matrix(bar))
#   a b
# w 1 5
# x 2 6
# y 3 7
# x 2 6
# z 4 8

Here is the problem: How to rbind() two data frames, having duplicated rows (with the same row name) removed?

Duplicate row names are not allowed in data.frames. – nograpes Oct 23 '12 at 20:11 — nograpes, Oct 23 '12 at 20:11

Ben Bolker · Accepted Answer · 2012-10-23T20:35:46.177

6

How about

duprows <- which(!is.na(match(rownames(bar),rownames(foo))))
rbind(foo,bar[-duprows,])

?

Or (based on comments below)

duprows <- rownames(bar) %in% rownames(foo)
rbind(foo, bar[!duprows,])

Several variations are possible depending on (1) selected matched or unmatched; (2) finding numeric or logical values for the matches.

edited Oct 23 '12 at 20:35

answered Oct 23 '12 at 20:13

Ben Bolker

211,554
25
370
453

1

Nice! But is there any shorter solution? I like to think at the R problems as creative puzzles! – Ali Oct 23 '12 at 20:20
rbind(foo, bar[which(is.na(match(rownames(bar), rownames(foo)))),]) – Ali Oct 23 '12 at 20:26
2

@AliSharifi -- `rbind(foo, bar[!rownames(bar) %in% rownames(foo),])` feels most natural to *me*. – Josh O'Brien Oct 23 '12 at 20:29
@JoshO'Brien Perfect! Why don't you make it as an answer? – Ali Oct 23 '12 at 20:31
@mplourde Very nice idea. It can be used with setNames(). – Ali Oct 23 '12 at 20:34
`setdiff(rownames(bar), rownames(foo))` is one more of the numerous variations Ben alluded to... – Josh O'Brien Oct 23 '12 at 21:33

score 3 · Answer 2 · answered Sep 16 '19 at 17:51

3

If I understand correctly what you're after, a nice, concise way to do this in one line of code:

unique(rbind(bar, foo))
#   a b
# x 2 6
# z 4 8
# w 1 5
# y 3 7

answered Sep 16 '19 at 17:51

Jack

31
4

concise but *slightly* less efficient (since it binds the data frames and *then* throws away the duplicates) – Ben Bolker Jul 08 '21 at 21:19

rbind data frames, duplicated rownames issue

2 Answers2

Linked