Gather ragged data frame into key-value columns - r

Question

I recently discovered how to create ragged data frames using the I function, but are having a hard time integrating them with tidyr, ggplot2 and the rest of the Hadleyverse. More specifically, how do you gather a column containing named vectors into key-value-columns?
Suppose I create a data frame like this
make.vector <- function(length.out){
x <- sample(9, length.out)
names(x) <- switch(length.out,
"Alice",
c("Bob", "Charlie"),
c("Dave", "Erin", "Frank"),
c("Gwen", "Harold", "Inez", "James"))
x
}
mydf <- data.frame(Game = gl(3, 3, labels=LETTERS[1:3]),
Set = rep(1:3, 3),
Score = I(lapply(rep(2:4, each=3), make.vector)))
producing
> print(mydf)
Game Set Score
1 A 1 8, 3
2 A 2 2, 8
3 A 3 3, 8
4 B 1 1, 5, 4
5 B 2 2, 3, 5
6 B 3 2, 8, 5
7 C 1 7, 2, 3, 4
8 C 2 1, 6, 3, 7
9 C 3 6, 9, 3, 7
The data frame can be manipulated with dplyr and tidyr in a straight forward manner as long as the results are of the expected length.
mydf %>%
mutate(nPlayers = sapply(Score, length))
mydf %>%
group_by(Game) %>%
summarize(TotalScore = list(Reduce("+", Score)))
However, I cannot figure out how to create multiple rows of result for each original row. Suppose I want to create the following data frame by manipulating mydf:
Game Set Player Score
1 A 1 Bob 8
2 A 1 Charlie 3
3 A 2 Bob 2
4 A 2 Charlie 8
5 A 3 Bob 3
6 A 3 Charlie 8
7 B 1 Dave 1
8 B 1 Erin 5
9 B 1 Frank 4
10 B 2 Dave 2
...
The only tool I know for doing so would be the gather function of the tidyr package, but it doesn't seem to play very well with non-atomic data.
mydf %>%
mutate(Player = lapply(Score, names)) %>%
gather(P = Player, S = Score)
I guess I could hack together a solution (as done in similar previous questions [1][2]),
cbind(
mydf[rep(1:nrow(mydf), sapply(mydf$Score, length)),
c("Game", "Set")],
data.frame(
Player = unlist(lapply(mydf$Score, names)),
Score = unlist(mydf$Score)
)
)
but I have a feeling I will have a hard time digesting it if look back at the code next week. Is there a "official" or at least smarter way to do this? Otherwise I'll make a general function for it and add to my personal library.
Update
In the light of David's answer below I figured out that the same result can be achieved with dplyr too.
mydf %>%
group_by(Game, Set) %>%
do(with(., data.frame(Player = names(unlist(Score)),
Score = unlist(Score))))
# Game Set Player Score
# 1 A 1 Bob 8
# 2 A 1 Charlie 6
# 3 A 2 Bob 7
# 4 A 2 Charlie 6
# 5 A 3 Bob 5
# 6 A 3 Charlie 8
# 7 B 1 Dave 1
# 8 B 1 Erin 9
# 9 B 1 Frank 3
# 10 B 2 Dave 8
# .. ... ... ... ...
# Warning message:
# In rbind_all(out[[1]]) : Unequal factor levels: coercing to character

Answer 1

I would try unlisting by group using data.table. You can run this only once per each group while storing it in a temporary variable using curly brackets (as you would do within a function) within the jth expression
library(data.table)
setDT(mydf)[, {
temp <- unlist(Score)
.(Player = names(temp), Score = temp)
}, by = .(Game, Set)]
# Game Set Player Score
# 1: A 1 Bob 2
# 2: A 1 Charlie 9
# 3: A 2 Bob 6
# 4: A 2 Charlie 3
# 5: A 3 Bob 2
# 6: A 3 Charlie 8
# 7: B 1 Dave 1
# 8: B 1 Erin 6
# 9: B 1 Frank 5
# 10: B 2 Dave 3
#...