Unconventional data frame reshaping

Question

I'm trying to reshape table from

col1| col2|col3|id
v1  | v2  | v3 |1
v2  | v1  | v3 |2

to

col1|col2|id
v1  | v2 | 1
v2  | v3 | 1
v2  | v1 | 2
v1  | v3 | 2

Basically create pair of following columns for each id.

Sorry for not clear title, but not sure how to describe what I'm trying to do.

Answer 1

for the case presented:

library(dplyr)
bind_rows(
    select(df, -col3),
    select(df,  col1=col2, col2=col3, id)
)

for more columns, assuming id is the last column:

library(dplyr)

NcolsToReduce <- ncol(df) - 2

lapply(1:NcolsToReduce, function(i){
  x <- select(df, i, i+1, id)
  names(x) <- c("col1", "col2", "id")
  x
  }) %>% bind_rows()

I think that'd work.

Answer 2

The solution below should be working for an arbitrary number of columns and an id variable which can be any column (not just the last one).

It uses melt() to reshape from wide to long format and shift() to get the next row as second column of the column pair:

library(data.table)
melt(setDT(df), id.vars = "id")[
  , .(col1 = head(value, -1L), col2 = head(shift(value, 1L, type = "lead"), -1L)), by = id]

    id col1 col2
1:  1   v1   v2
2:  1   v2   v3
3:  2   v2   v1
4:  2   v1   v3

Data

library(data.table)
df <- fread(
  "col3| col2|col1|id
v1  | v2  | v3 |1
v2  | v1  | v3 |2",
  sep = "|")