Question

When I say unexpected, I mean unexpected for me. Allow me to illustrate. We have the two data.frames:

b1<-data.frame(a=c("a","b"),b=1:2)
b2<-data.frame(a=c("a","b"),c=1:2)

Merge produces the following

> merge(b1,b2)
  a b c
1 a 1 1
2 b 2 2

But when we have the data.frames

b1<-data.frame(a=c("a","a"),b=1:2)
b2<-data.frame(a=c("a","a"),c=1:2)

merge produces the

> merge(b1,b2)
  a b c
1 a 1 1
2 a 1 2
3 a 2 1
4 a 2 2

when I expect

  a b c
  a 1 1
  a 2 2

Why the two different results?

Answer 1

This is by design. base merge uses match on columns specified (or not specified). In case 1, it found only a single match for each value of a so there were no duplicates. But in case 2, it found two matches:

> b1$a %in% b2$a 
[1] TRUE TRUE  

for each a and therefore returned all possible matches. See ?merge for more information. join in plyr has the option of matching only the first match.

> join(b1,b2, match="first")
Joining by: a
  a b c
1 a 1 1
2 a 2 1