R 查找（有序）组中以前出现的情态索引

用一个例子来解释更容易。考虑这些数据：

library(dplyr)
n_data       = 20
set.seed(123)
data_standout = data.frame(group = sample(c('kids','monkeys','banana', 'latte'),
                                         size = n_data, replace=TRUE), 
                          time  = runif(n_data),
                          modality = sample(c('grande','small','large'),
                                            size = n_data, replace=TRUE),
                          stringsAsFactors = FALSE)
data_standin  = data.frame(group = c('kids','monkeys','banana', 'latte'), 
                           time = runif(4, 0, min(data_standout$time)), 
                           modality = rep('small', 4), stringsAsFactors = FALSE)

data_final    = bind_rows(data_standout, data_standin)

嗯。现在让我们特别考虑一组：猴子。

# A tibble: 7 x 3
# Groups:   group [1]
    group       time modality
    <chr>      <dbl>    <chr>
1 monkeys 0.09211798    small
2 monkeys 0.17505265   grande
3 monkeys 0.32037324    large
4 monkeys 0.43489274    large
5 monkeys 0.46677904    small
6 monkeys 0.48861303   grande
7 monkeys 0.78229430    small

---

速度不是主要要求，但dplyr是（数据管线的其余部分在dplyr中）。

以

猴子

数据为例：

monkeys %>% 
    arrange(time) %>% 
    group_by(group) %>% 
    mutate(small_idx = lag(cummax(ifelse(modality == 'small', seq_along(modality), 0))))

# A tibble: 7 x 4
# Groups:   group [1]
#    group       time modality small_idx
#   <fctr>      <dbl>   <fctr>     <dbl>
#1 monkeys 0.09211798    small        NA
#2 monkeys 0.17505265   grande         1
#3 monkeys 0.32037324    large         1
#4 monkeys 0.43489274    large         1
#5 monkeys 0.46677904    small         1
#6 monkeys 0.48861303   grande         5
#7 monkeys 0.78229430    small         5

cummax

给出了迄今为止看到的

small

的最大索引：

cummax(with(monkeys, ifelse(modality == 'small', seq_along(modality), 0)))
# [1] 1 1 1 1 5 5 7

lag

给出了前面看到的

small

的最大索引：

lag(cummax(with(monkeys, ifelse(modality == 'small', seq_along(modality), 0))))
# [1] NA  1  1  1  1  5  5

---

这太酷了：这个技巧（ifelse+cumsum）可以很容易地推广到其他相关问题！万分感谢

cummax(with(monkeys, ifelse(modality == 'small', seq_along(modality), 0)))
# [1] 1 1 1 1 5 5 7

lag(cummax(with(monkeys, ifelse(modality == 'small', seq_along(modality), 0))))
# [1] NA  1  1  1  1  5  5