如何实现不受数据顺序影响的自动分组函数，或恢复原数据排序

背景 #

此前在Stack Overflow提出过相关分组问题，但数据场景复杂时原有方案失效。为此构建四类测试场景验证问题：

• sample_dat：新增by组类别
• sample_dat2：行随机排序
• sample_dat3：随机删除部分行
• sample_dat4：随机重复部分行

测试结论 #

仅数据顺序会导致原有方案出错（删除行无影响，重复行出现与sample_dat2相同的问题）。

排序尝试失败 #

尝试通过sample_dat2_to1 <- sample_dat2[order(cat, year, type),]将sample_dat2还原为sample_dat的排序，但未达到预期，输出示例：

A B C  D   cat   type observations year
 1: 1 0 2 NA cat X type 1            1 2010
 2: 1 0 2 NA cat X type 1            1 2010
 3: 1 0 2  4 cat X type 1            1 2010
 4: 3 4 3  1 cat X type 2            3 2010
 5: 3 4 3  1 cat X type 2            3 2010
 6: 3 4 3  1 cat X type 2            3 2010
 7: 1 0 2  2 cat X type 3            3 2010
 8: 1 0 2  2 cat X type 3            3 2010

核心诉求 #

如何将sample_dat2还原为sample_dat的排序以适配原有函数，或修改Waldi与ThomasIsCoding编写的observations_grp函数，使其独立于数据顺序？

测试数据与代码 #

sample_dat <- structure(list(A = c(1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 
1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 
3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3, 1, 3), B = c(0, 4, 0, 4, 0, 
4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 
0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 4, 0, 
4), C = c(2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 
3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 
2, 3, 2, 3, 2, 3, 2, 3, 2, 3), D = c(NA, 1, 2, 0, NA, NA, 2, 
35, NA, 1, 2, NA, NA, NA, 2, 1, 4, 1, 2, 2, NA, NA, 2, 2, NA, 
1, 2, 0, NA, NA, 2, 35, NA, 1, 2, NA, NA, NA, 2, 1, 4, 1, 2, 
2, NA, NA, 2, 2), cat = c("cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y", "cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y", "cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y", "cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y", "cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y", "cat X", "cat X", "cat X", "cat X", 
"cat Y", "cat Y", "cat Y", "cat Y"), type = c("type 1", "type 2", "type 3", "type 4", 
"type 1", "type 2", "type 3", "type 4", "type 1", "type 2", "type 3", 
"type 4", "type 1", "type 2", "type 3", "type 4", "type 1", "type 2", 
"type 3", "type 4", "type 1", "type 2", "type 3", "type 4", "type 1", 
"type 2", "type 3", "type 4", "type 1", "type 2", "type 3", "type 4", 
"type 1", "type 2", "type 3", "type 4", "type 1", "type 2", "type 3", 
"type 4", "type 1", "type 2", "type 3", "type 4", "type 1", "type 2", 
"type 3", "type 4"), observations = c(1, 3, 3, 2, 0, 0, 3, 3, 
1, 3, 3, 2, 0, 0, 3, 3, 1, 3, 3, 2, 0, 0, 3, 3, 1, 3, 3, 2, 0, 
0, 3, 3, 1, 3, 3, 2, 0, 0, 3, 3, 1, 3, 3, 2, 0, 0, 3, 3), year = c(2010, 
2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 
2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 2010, 
2010, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 
2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 
2020, 2020, 2020)), class = c("tbl_df", "tbl", "data.frame"), row.names = c(NA, 
-48L))

sample_dat <- data.table(sample_dat)
set.seed(1)
sample_dat2 <- sample_dat[sample(nrow(sample_dat)),]
# row_nums <- floor(runif(10, min=1, max=48))
row_nums <- c(43, 14, 22, 16, 31, 13, 23, 37, 4, 42)
sample_dat3 <- sample_dat[-row_nums]
sample_dat4 <- rbind(sample_dat, sample_dat[row_nums])

observations_grp <- function(x) {
  cumsum_i <- 0
  nxtgrp <-  F
  n <- length(x)
  grp <- rep(0,n)
  grp_i <- 0;
  for (i in 1:n) {
    if (nxtgrp) {grp_i <- grp_i + 1; cumsum_i <- 0;}
    nxtgrp <- !((cumsum_i + x[i]) < 2)
    cumsum_i <- cumsum_i + x[i]
    grp[i] <- grp_i
  }
  grp
}

sample_dat[,`:=`(type = last(type), observations=sum(observations)),
        .(cat, year ,observations_grp(observations))
][]

sample_dat2[,`:=`(type = last(type), observations=sum(observations)),
        .(cat, year ,observations_grp(observations))
][]

sample_dat3[,`:=`(type = last(type), observations=sum(observations)),
        .(cat, year ,observations_grp(observations))
][]


sample_dat4[,`:=`(type = last(type), observations=sum(observations)),
        .(cat, year ,observations_grp(observations))
][]

内容的提问来源于stack exchange，提问作者Tom