CODEWITHSUNDEEP
rmerge
Rmeow
Rmeow

Reputation: 113

Merging datasets where the key is on the column of one dataset and row of another in R

How do I merge datasets where the key is on the column of one dataset and row of another in R?

Example data:

group = c("a", "b", "c", "c")
id1 = c(1, 0, 0, 0) 
id2 = c(0, 1, 0, 1)
id3 = c(0, 0, 1, 0) 
df1 = data.frame(group,id1, id2, id3) 
df1

id = c("id1", "id2", "id3") 
iv1 = c(2, 3, 3) 
iv2 = c(3, 2, 3) 
df2 = data.frame(id, iv1, iv2) 
df2

I like to merge these two datasets by ids. I can do it via ifelse:

df1$iv1 = ifelse(df1$id1 == 1, 2, 3)
df1$iv2 = ifelse(df1$id2 == 1, 2, 3)

with the result:

  group id1 id2 id3 iv1 iv2
1     a   1   0   0   2   3
2     b   0   1   0   3   2
3     c   0   0   1   3   3
4     c   0   1   0   3   2

The ifelse way would be tedious if say, I have 1000 variables to merge. Is there a more efficient way to merge datasets like these?

Upvotes: 1

Views: 94

Answers (3)

slushy
slushy

Reputation: 3377

Both Pierre and Istrel's answers are good. For this and more complex manipulations, you could also use Hadley Wickham's popular tidyr package:

install.packages('tidyr', repos='http://cran.rstudio.org')
library(tidyr)

g1 <- gather(df1, idx, id_val, -group)  # colnames are in 'idx'; 12 rows total
g1 <- g1[g1$id_val==1, ]                # drop rows with id_val == 0
g2 <- merge(g1, df2, by.x='idx', by.y='id')
g3 <- spread(g2, idx, id_val)           # pivot the 'idx' column back out
g3
#   group iv1 iv2 id1 id2 id3
# 1     a   2   3   1  NA  NA
# 2     b   3   2  NA   1  NA
# 3     c   3   2  NA   1  NA
# 4     c   3   3  NA  NA   1
g3[is.na(g3)] <- 0
g3
#   group iv1 iv2 id1 id2 id3
# 1     a   2   3   1   0   0
# 2     b   3   2   0   1   0
# 3     c   3   2   0   1   0
# 4     c   3   3   0   0   1

Upvotes: 0

Istrel
Istrel

Reputation: 2588

Pierre Lafortune have already given a good answer. Still I will post my solution:

ids <- colnames(df1[, 2:4])
ids <- apply(df1[, 2:4], 1, function(x) return(ids[as.logical(x)]))

df1$id <- ids
new_df <- merge(df1, df2, by="id", all.x = TRUE, sort=FALSE)

> new_df
   id group id1 id2 id3 iv1 iv2
1 id1     a   1   0   0   2   3
2 id2     b   0   1   0   3   2
3 id2     c   0   1   0   3   2
4 id3     c   0   0   1   3   3

Upvotes: 0

Pierre L
Pierre L

Reputation: 28461

We can use the fact that the column order of the first matches the row order of the second.

df1[c("iv1", "iv2")] <- df2[max.col(df1[-1]),-1]
#   group id1 id2 id3 iv1 iv2
# 1     a   1   0   0   2   3
# 2     b   0   1   0   3   2
# 3     c   0   0   1   3   3
# 4     c   0   1   0   3   2

Upvotes: 4

Related Questions