CODEWITHSUNDEEP
rdataframemerge
MeC
MeC

Reputation: 463

How to merge dataframes based on *partial* row overlap?

I would like to merge dataframes in R such that only only those observations whose rows partially correspond across the dataframes are kept.

I have two dataframes (These are toy dataframes - the actual ones have hundreds of columns.):

    V1             V2      V3 
    rabbit         001     M
    squirrel       001     M
    cow            001     M
    rabbit         004     M
    squirrel       004     M
    skunk          004     M

    V1             V2       V3
    rabbit         001      B
    squirrel       001      B
    skunk          001      B
    rabbit         004      B
    squirrel       004      B
    skunk          008      B

Desired outcome:

    V1             V2       V3
    rabbit         001      M
    squirrel       001      M
    rabbit         004      M
    squirrel       004      M
    rabbit         001      B
    squirrel       001      B
    rabbit         004      B
    squirrel       004      B

merge and dplyr::inter_join aren't quite the right functions for this. What is?

Upvotes: 1

Views: 225

Answers (2)

divibisan
divibisan

Reputation: 12155

d.b's answer is likely much more efficient, but if you prefer to think about the problem in terms of JOIN operations, you can do this with 3 dplyr join operations:

library(dplyr)

# Perform an inner_join with just the columns that you want to match
match_rows <- inner_join(df1[,1:2], df2[,1:2])
match_rows

        V1 V2
1   rabbit  1
2 squirrel  1
3   rabbit  4
4 squirrel  4

# Then left_join that with each dataframe to get the matching rows from each
#  and then bind them together as rows
bind_rows(left_join(match_rows, df1),
          left_join(match_rows, df2))

        V1 V2 V3
1   rabbit  1  M
2 squirrel  1  M
3   rabbit  4  M
4 squirrel  4  M
5   rabbit  1  B
6 squirrel  1  B
7   rabbit  4  B
8 squirrel  4  B

Upvotes: 1

d.b
d.b

Reputation: 32538

rbind(d1, d2)[ave(1:(nrow(d1) + nrow(d2)),
           Reduce(paste, rbind(d1, d2)[c("V1", "V2")]),
           FUN = length) > 1,]
#         V1 V2 V3
#1    rabbit  1  M
#2  squirrel  1  M
#4    rabbit  4  M
#5  squirrel  4  M
#7    rabbit  1  B
#8  squirrel  1  B
#10   rabbit  4  B
#11 squirrel  4  B

Data

#dput(d1)
structure(list(V1 = c("rabbit", "squirrel", "cow", "rabbit", 
"squirrel", "skunk"), V2 = c(1L, 1L, 1L, 4L, 4L, 4L), V3 = c("M", 
"M", "M", "M", "M", "M")), row.names = c(NA, 6L), class = "data.frame")

#dput(d2)
structure(list(V1 = c("rabbit", "squirrel", "skunk", "rabbit", 
"squirrel", "skunk"), V2 = c(1L, 1L, 1L, 4L, 4L, 8L), V3 = c("B", 
"B", "B", "B", "B", "B")), row.names = 7:12, class = "data.frame")

Upvotes: 2

Related Questions