>我有以下数据集：
structure(list(ID = c(5L, 6L, 7L, 8L, 10L), chain = c("x49",
"x43", "x32 > x42 > x49 > x45 > x20 > x50 > x38", "x54 > x44",
"x38 > x38")), row.names = c(NA, -5L), class = c("data.table",
"data.frame"))
ID chain
1: 5 x49
2: 6 x43
3: 7 x32 > x42 > x49 > x45 > x20 > x50 > x38
4: 8 x54 > x44
5: 10 x38 > x38
链列代表产品的购买过程，也缺少一些信息(开始和购买(。目标是对链中的每个值进行两次计数(起点，例如从和目的地，例如到( 为了能够做到这一点，我需要重构数据集。例如，重组后的链x54 > x44应该是这样的：
from to
1 start x54
2 x54 x44
3 x44 buy
整个结果应如下所示：
from to
1 start x49
2 x49 buy
3 start x43
4 x43 buy
5 start x32
6 x32 x42
7 x42 x49
8 x49 x45
9 x45 x20
10 x20 x50
11 x38 buy
12 start x54
13 x54 x44
14 x44 buy
15 start x54
16 x54 x44
17 x44 buy
18 start x38
19 x38 x38
20 x38 buy
我已经尝试过了，但我不确定这是否是一个好主意(也不知道如何在这里继续(。
df <- strsplit(df$chain, ">")
lapply(df, trimws)
性能可能很重要，因为链可能会变得很长(30 个项目(，并且整个数据集有 100k 行。
一种基本的 R 方法是拆分" > "上的字符串并创建一个组合所有值的数据帧。
do.call(rbind, lapply(strsplit(df$chain, " > "), function(x)
data.frame(from = c("start",x), to = c(x, "buy"))))
# from to
#1 start x49
#2 x49 buy
#3 start x43
#4 x43 buy
#5 start x32
#6 x32 x42
#7 x42 x49
#8 x49 x45
#9 x45 x20
#10 x20 x50
#11 x50 x38
#12 x38 buy
#13 start x54
#14 x54 x44
#15 x44 buy
#16 start x38
#17 x38 x38
#18 x38 buy
使用类似的方法tidyverse方法是
library(tidyverse)
map_dfr(str_split(df$chain, " > "), ~tibble(from = c("start",.), to = c(., "buy")))
我们可以将字符串粘贴在开头和结尾 str_c ，使用 separate_rows 用 tidyverse 扩展数据集
library(tidyverse)
dt %>%
mutate(chain = str_c("start > ", chain, " > buy")) %>%
separate_rows(chain) %>% group_by(ID) %>%
transmute(from = chain, to = lead(chain)) %>%
na.omit %>%
ungroup %>%
select(-ID)
# A tibble: 18 x 2
# from to
# <chr> <chr>
# 1 start x49
# 2 x49 buy
# 3 start x43
# 4 x43 buy
# 5 start x32
# 6 x32 x42
# 7 x42 x49
# 8 x49 x45
# 9 x45 x20
#10 x20 x50
#11 x50 x38
#12 x38 buy
#13 start x54
#14 x54 x44
#15 x44 buy
#16 start x38
#17 x38 x38
#18 x38 buy