Ari B. Friedman Ari B. Friedman - 1 month ago 10
R Question

Adding rows in `dplyr` output

In traditional

plyr
, returned rows are added automagically to the output even if they exceed the number of input rows for that grouping:

set.seed(1)
dat <- data.frame(x=runif(10),g=rep(letters[1:5],each=2))
> ddply( dat, .(g), function(df) df[c(1,1,1,2),] )
x g
1 0.26550866 a
2 0.26550866 a
3 0.26550866 a
4 0.37212390 a
5 0.57285336 b
6 0.57285336 b
7 0.57285336 b
8 0.90820779 b
9 0.20168193 c
10 0.20168193 c
11 0.20168193 c
12 0.89838968 c
13 0.94467527 d
14 0.94467527 d
15 0.94467527 d
16 0.66079779 d
17 0.62911404 e
18 0.62911404 e
19 0.62911404 e
20 0.06178627 e


I cannot figure out how to do the same in
dplyr
. Some attempts:

dat %>% group_by(g) %>% summarise( xbar = mean(x) )

> dat %>% group_by(g) %>% summarise( xbar = runif(3) )
Error: expecting a single value

# Getting creative...

> dat %>% group_by(g) %>% function(x) x[c(1,1,1,2),]

# Nope.


How do I do this?

The specific use case I'm butting up against is splitting a
\n
-delimited text field and making it "long," but I use this feature of
ddply
all the time for many purposes.

Answer

Try this:

 dat %>% 
     group_by( g ) %>% 
     do( .[c(1,1,1,2), ] ) %>% 
     ungroup()