如何以R“ data.table”方式记录非重叠范围（或时间间隔）？

Question

我的目标是在每个ID具有多个可能重叠的曝光范围时，为每个ID定义唯一的非重叠范围（或时间间隔）。我发现R“ IntervalSurgeon”包中的“ flatten”函数可以实现任务。我的问题是：如何以“ data.table”方式有效地执行相同的任务并获得相同的“ tab_out”输出？

library(data.table)

library(IntervalSurgeon)

set.seed(2019)

N <- 3 # number of IDs

IDs <- paste0("ID", 1:N) # unique IDs

K <- 4 # number of exposures per ID

DT <- data.table(IDs = rep(IDs, each = K), 
    starts = sample(1:20, N * K, replace = T))[,
    ends := starts + sample(1:5, N * K, replace = T)]

DT <- DT[order(IDs, starts),]

tab_out <- DT[, as.list(data.table(
    flatten(as.matrix(cbind(starts, ends))))), 
    by = IDs]

DT
    IDs starts ends
 1: ID1      7   11
 2: ID1     13   17
 3: ID1     15   16
 4: ID1     16   18
 5: ID2      1    5
 6: ID2      1    4
 7: ID2      2    3
 8: ID2     17   19
 9: ID3      3    6
10: ID3     13   16
11: ID3     14   15
12: ID3     16   21

tab_out
   IDs V1 V2
1: ID1  7 11
2: ID1 13 18
3: ID2  1  5
4: ID2 17 19
5: ID3  3  6
6: ID3 13 21

Answer 1

**样本数据**

library( data.table )
library( intervals )

DT <- fread("
IDs starts ends
ID1      7   11
ID1     13   17
ID1     15   16
ID1     16   18
ID2      1    5
ID2      1    4
ID2      2    3
ID2     17   19
ID3      3    6
ID3     13   16
ID3     14   15
ID3     16   21")

code
myfun <- function( y ) {
  data.table::as.data.table( 
    intervals::interval_union(
      intervals::Intervals( as.matrix( y ) ), check_valid = TRUE ) 
    )
}

DT[, myfun( .SD ), by = .(IDs)]

#    IDs V1 V2
# 1: ID1  7 11
# 2: ID1 13 18
# 3: ID2  1  5
# 4: ID2 17 19
# 5: ID3  3  6
# 6: ID3 13 21