R - 根据数据框中的条件按组设置值

Question

我有以下数据集。

group	value	row_name	overlaps
group_a	4	1	2
group_a	5	2	3, 5
group_a	48	3	4, 5
group_a	54	4	5
group_a	12	5
group_b	12	6	7
group_b	1	7

重叠列指示哪些行具有一定的“重叠”。

我希望，仅对于值大于 10 的行，将相应“重叠”列中指示的所有行的值列中的数据替换为零。

预期输出：

group	value	row_name	overlaps
group_a	4	1	2
group_a	5	2	3, 5
group_a	48	3	4, 5
group_a	0	4	5
group_a	0	5
group_b	12	6	7
group_b	0	7

可重现的例子：

data <- data.frame(group = c("group_a", "group_a", "group_a", "group_a",
                             "group_a", "group_b", "group_b"),
                   value = c(4, 5 , 48, 54, 12, 12, 1),
                   row_name = c("1", "2", "3", "4", "5", "6", "7"),
                   overlaps = c("2", "3, 5", "4, 5", "5", "", "7", ""))

我不知道这是否是一个非常复杂的问题，但我被困了几个小时，我不知道如何解决它。

有人对我如何使用 dplyr 或 data.table 按组解决此问题有任何建议吗？

Answer 1

另一种解决方案，基于tidyverse：

library(tidyverse)

data <- data.frame(group = c("group_a", "group_a", "group_a", "group_a",
                             "group_a", "group_b", "group_b"),
                   value = c(4, 5 , 48, 54, 12, 12, 1),
                   row_name = c("1", "2", "3", "4", "5", "6", "7"),
                   overlaps = c("2", "3, 5", "4, 5", "5", "", "7", ""))

data %>% 
  separate(
    overlaps, into=c("o1", "o2"), sep=", ", fill="right", remove=F) %>% 
  mutate(across(o1:o2, ~ ifelse(value > 10, get(cur_column()), 0)),
         value = ifelse(row_number() %in% c_across(o1:o2), 0, value)) %>% 
  select(-o1, -o2)

#>     group value row_name overlaps
#> 1 group_a     4        1        2
#> 2 group_a     5        2     3, 5
#> 3 group_a    48        3     4, 5
#> 4 group_a     0        4        5
#> 5 group_a     0        5         
#> 6 group_b    12        6        7
#> 7 group_b     0        7

Answer 2

strsplit overlaps 列的子集，仅是 data$value > 10 的子集，然后使用 row_names 的不同集合用 0 覆盖原始数据：

gr10 <- data$value > 10
sel <- Map(paste, data$group[gr10], strsplit(data$overlaps, ",\\s+")[gr10], sep="|")
sel <- Reduce(union, sel)
sel
#[1] "group_a|4" "group_a|5" "group_a|"  "group_b|7"
data$value[do.call(paste, c(data[c("group","row_name")], sep="|")) %in% sel] <- 0
data
#    group value row_name overlaps
#1 group_a     4        1        2
#2 group_a     5        2     3, 5
#3 group_a    48        3     4, 5
#4 group_a     0        4        5
#5 group_a     0        5         
#6 group_b    12        6        7
#7 group_b     0        7         

如果row_names 在整个数据集中是唯一的，您可以使用更简单的逻辑：

sel <- Reduce(union, strsplit(data$overlaps, ",\\s+")[data$value > 10])
sel
#[1] "4" "5" "7"
data$value[data$row_name %in% sel] <- 0

奖励 data.table 解决方案：

library(data.table)
setDT(data)

data[
  data[value > 10, .(row_name=unlist(strsplit(overlaps, ",\\s+"))), by=group],
  on=.(group, row_name),
  value := 0
]