multiple rows in one after join - r

Question

i have to join two dataframes. the joining is working properly, my only problem is:
sometimes i have 2 rows in a dataset, that have to be joined with the other dataset. after the joining the rows are displayed directly one below the other. I need them in one row:
here a small example:
df1 <- data.frame(A = c(1:5),
B = c(12, 13, 14, 15, 16),
C = c("a", "b", "c", "d", "e"))
df2 <- data.frame(A = c(1,1,2,2,3),
E = c(112, 145, 546, 674, 287),
J = c("t", "e", "v", "p", "s"))
merged <- inner_join(df1, df2, by = "A")
A | B | c | E | J |
---------------------
1 | 12| a | 112 | t
1 | 12| a | 145 | e
2 | 13| b | 546 | v
2 | 13| b | 674 | p
3 | 14| c | 287 | s
i need it this way:
A | B | c | E.x | J.x | E.y | J.y
--------------------------------
1 | 12| a | 112 | t | 145 | e
2 | 13| b | 546 | v | 674 | p
all joined rows together in one row in the new dataset
thank you for your help
derlu
**update:**
My original data frames have
354 items with 1535 observation
246 items 203 observation
They are merged by the first column id_merge. Sometimes I have 2 matches, sometimes 3 identical „id_merge“ values.

Answer 1

One approach is to split up the dataframe by occurrence of each key, and then do another join. First, we create a key (n) and split up the dataframe:
merged <- merged %>%
group_by(A, B, C) %>%
mutate(n = 1:n())
# A B C E J n
# <dbl> <dbl> <fct> <dbl> <fct> <int>
# 1 1 12 a 112 t 1
# 2 1 12 a 145 e 2
# 3 2 13 b 546 v 1
# 4 2 13 b 674 p 2
# 5 3 14 c 287 s 1
merged_list <- split(merged, merged$n) %>%
map(select, -n)
# $`1`
# # A tibble: 3 x 5
# # Groups: A [3]
# A B C E J
# <dbl> <dbl> <fct> <dbl> <fct>
# 1 1 12 a 112 t
# 2 2 13 b 546 v
# 3 3 14 c 287 s
# $`2`
# # A tibble: 2 x 5
# # Groups: A [2]
# A B C E J
# <dbl> <dbl> <fct> <dbl> <fct>
# 1 1 12 a 145 e
# 2 2 13 b 674 p
Then use Reduce to join the dataframes:
Reduce(function(x, y) left_join(x, y, by = c("A", "B", "C")),
merged_list)
# A B C E.x J.x E.y J.y
# <dbl> <dbl> <fct> <dbl> <fct> <dbl> <fct>
# 1 1 12 a 112 t 145 e
# 2 2 13 b 546 v 674 p
# 3 3 14 c 287 s NA NA

Answer 2

It's not cool these days, but Base R's reshape actually shines for this as it deals with unequal group sizes and labelling all in one go. You need to add a time variable first though to label each group:
merged <- merge(df1, cbind(df2, time=ave(df2$A, df2$A, FUN=seq_along)), by="A")
reshape(merged, idvar=names(df1), direction="wide", sep="_")
# A B C E_1 J_1 E_2 J_2
#1 1 12 a 112 t 145 e
#3 2 13 b 546 v 674 p
#5 3 14 c 287 s NA <NA>
It actually fits in a dplyr pipeline not too bad:
df2 %>%
group_by(A) %>%
mutate(time=row_number()) %>%
inner_join(df1,.,by="A") %>%
reshape(idvar=names(df1), direction="wide", sep="_")

Answer 3

With data.table, you can split each group of (E, J) rows having the same (A, B, C) values by row, and cbind that list together. Here .I is the row number, .N is the number of rows in the group, and .SD is the group of rows.
This only works if the number of repeated rows is the same in all groups (except for groups with 1 row)
library(data.table)
setDT(merged)
merged[, if(.N > 1) do.call(cbind, split(.SD, .I))
, by = .(A, B, C)]
# A B C 1.E 1.J 2.E 2.J
# 1: 1 12 a 112 t 145 e
# 2: 2 13 b 546 v 674 p

Answer 4

It will be like this and applicable for larger datasets as well
df1 <- data.frame(A = c(1:5),
B = c(12, 13, 14, 15, 16),
C = c("a", "b", "c", "d", "e"))
df2 <- data.frame(A = c(1,1,2,2,3),
E = c(112, 145, 546, 674, 287),
J = c("t", "e", "v", "p", "s"))
df3 <- merge(df1, df2, by=c("A"))
df4 <- data.frame()
for (j in 1:(nrow(df3)-1))
for (k in (j+1):nrow(df3))
for (i in 2:ncol(df3))
if(df3[j,1] == df3[k,1] && all(df3[j,i:ncol(df3)]!=df3[k,i:ncol(df3)])){
if(nrow(df4) == 0)
df4 <- data.frame(c(df3[j,],df3[k,i:ncol(df3)]))
else
df4 <- rbind(df4,c(df3[j,],df3[k,i:ncol(df3)]))
break
}