卷起一个 data.table

Question

我有一个数据表:

> (mydt <- data.table(id=c(1,1,1,1,2,2),
                      time=1:6,
                      v1=letters[1:6],
                      v2=LETTERS[1:6],
                      key=c("id","time")))
   id time v1 v2
1:  1    1  a  A
2:  1    2  b  B
3:  1    3  c  C
4:  1    4  d  D
5:  2    5  e  E
6:  2    6  f  F

我想将它“卷起来”(这是正确的术语吗？)，到“更改”表:
对象 1更改了 3 次(从时间戳 1 到 2、2 到 3、3 到 4)
对象 2改变一次(时间5到6)；
我对初始 v1 感兴趣最后 v2 .
所以，结果应该是:

> (res <- data.table(beg.time=c(1,2,3,5),
                     end.time=c(2,3,4,6),
                     v1=c('a','b','c','e'),
                     v2=c('B','C','D','F'),
                     key=c("beg.time","end.time")))
   beg.time end.time v1 v2
1:        1        2  a  B
2:        2        3  b  C
3:        3        4  c  D
4:        5        6  e  F

Answer 1

感谢您提供可重现的示例!这是一个尝试。

首先，请注意，您可以使用以下头尾习语将相距一定距离的向量的条目彼此相邻放置:

x <- letters[1:5]
cbind(head(x, -1), tail(x, -1))
     # [,1] [,2]
# [1,] "a"  "b" 
# [2,] "b"  "c" 
# [3,] "c"  "d" 
# [4,] "d"  "e" 
cbind(head(x, -2), tail(x, -2))
     # [,1] [,2]
# [1,] "a"  "c" 
# [2,] "b"  "d" 
# [3,] "c"  "e" 

然后，我们可以使用 by data.table 的功能按组执行此操作。

mydt[,{
    ## if there's just one row in the group of ID's, return nothing
    if (.N == 1) return(NULL) 
    else {
        list(
            ## head and tail take the first and last parts of a vector
            ## this will place an element next to its subsequent element
            beg.time = head(time, -1),
            end.time = tail(time, -1),
            v1 = head(v1, -1),
            v2 = tail(v2, -1)
## group by ID
)}}, by = id]

#    id beg.time end.time v1 v2
# 1:  1        1        2  a  B
# 2:  1        2        3  b  C
# 3:  1        3        4  c  D
# 4:  2        5        6  e  F