将长形改成宽形并保留重复的行

Question

对于给定的数据集，我想将我的数据集从长格式转换为宽格式。我已经使用重塑功能来做到这一点。

id  status      timestamp   
1   assigned   2017-01-02  
1   done       2017-01-03  
1   locked     2017-01-04   
2   assigned   2017-01-02   
2   done       2017-01-03  
2   assigned   2017-01-03  
2   done       2017-01-04 
2   locked     2017-01-05  
3   assigned   2017-01-02  
3   done       2017-01-03 
3   locked     2017-01-04 
...

# reshape function to convert long format to Wide.
temp <- reshape(temp, idvar = "id", timevar = "status", direction = "wide")

结果：

id timestamp.assigned timestamp.done timestamp.locked
1 2017-01-02 2017-01-03 2017-01-04
2 2017-01-02 2017-01-03 2017-01-05
3 2017-01-02 2017-01-03 2017-01-04

当我这样做时，它会删除一些行，例如：对于 id 2，status=assigned 有多行匹配，它占用第一行。

如何在不删除行的情况下转换为宽。基本上，我不想丢失任何数据。

预期结果：
id timestamp.assigned timestamp.done timestamp.locked
1 2017-01-02 2017-01-03 2017-01-04
2 2017-01-02 2017-01-03 2017-01-05
2 2017-01-03 2017-01-04 2017-01-05
3 2017-01-02 2017-01-03 2017-01-04

或

id timestamp.assigned timestamp.done timestamp.locked
1 2017-01-02 2017-01-03 2017-01-04
2 2017-01-02 2017-01-03 NA
2 2017-01-03 2017-01-04 2017-01-05
3 2017-01-02 2017-01-03 2017-01-04

Answer 1

1。 cumsum()

Esther's approach 为每个新任务编号是可行的方法。

但是，R 已经有 cumsum() 函数可以用于此目的：

temp$key <- cumsum(temp$status == "assigned")
reshape(temp, idvar = c("key", "id"), timevar = "status", direction = "wide")

   id key timestamp.assigned timestamp.done timestamp.locked
1:  1   1         2017-01-02     2017-01-03       2017-01-04
2:  2   2         2017-01-02     2017-01-03             <NA>
3:  2   3         2017-01-03     2017-01-04       2017-01-05
4:  3   4         2017-01-02     2017-01-03       2017-01-04

2。分组cumsum()

尽管这解决了 OP 的原始问题，key 只是在 all ids 中编号 all 分配。如果 OP 希望为每个 id 单独编号分配，我们需要应用按 id 分组的 cumsum()。

实现此目的的一种方法是使用data.table 语法：

library(data.table)
setDT(temp)[, key := cumsum(status == "assigned"), by = id]
dcast(temp, id + key ~ status, value.var = "timestamp")

   id key   assigned       done     locked
1:  1   1 2017-01-02 2017-01-03 2017-01-04
2:  2   1 2017-01-02 2017-01-03       <NA>
3:  2   2 2017-01-03 2017-01-04 2017-01-05
4:  3   1 2017-01-02 2017-01-03 2017-01-04

dcast() 是基本 R 的 reshape(..., direction = "wide") 函数的替代品，该函数可从 reshape2 和 data.table 包中获得。

3。分组cumsum() 即时

data.table的dcast()的公式接口也接受表达式。有了这个，没有必要修改temp，通过在整形之前添加key 列之前。相反，这可以在 重塑时即时完成：

dcast(temp, id + ave(key <- status == "assigned", id, FUN = cumsum) ~ 
        paste0("timestamp.", status))

   id key timestamp.assigned timestamp.done timestamp.locked
1:  1   1         2017-01-02     2017-01-03       2017-01-04
2:  2   1         2017-01-02     2017-01-03             <NA>
3:  2   2         2017-01-03     2017-01-04       2017-01-05
4:  3   1         2017-01-02     2017-01-03       2017-01-04

数据

library(data.table)
temp <- fread(
  "id  status      timestamp   
1   assigned   2017-01-02  
1   done       2017-01-03  
1   locked     2017-01-04   
2   assigned   2017-01-02   
2   done       2017-01-03  
2   assigned   2017-01-03  
2   done       2017-01-04 
2   locked     2017-01-05  
3   assigned   2017-01-02  
3   done       2017-01-03 
3   locked     2017-01-04 ")

Answer 2

您可以做的一件事是添加一个变量，该变量为每个新分配提供一个唯一值。然后你可以用它来塑造你的数据

i <- 0

temp$key <- sapply(temp$status, function(x) {
  if(x == "assigned") {i <<- i+1; i}
  else {i}
})

temp

   id   status  timestamp key
1   1 assigned 2017-01-02   1
2   1     done 2017-01-03   1
3   1   locked 2017-01-04   1
4   2 assigned 2017-01-02   2
5   2     done 2017-01-03   2
6   2 assigned 2017-01-03   3
7   2     done 2017-01-04   3
8   2   locked 2017-01-05   3
9   3 assigned 2017-01-02   4
10  3     done 2017-01-03   4
11  3   locked 2017-01-04   4

temp2 <- reshape(temp, idvar = c("key", "id"), timevar = "status", direction = "wide")

temp2

  id key timestamp.assigned timestamp.done timestamp.locked
1  1   1         2017-01-02     2017-01-03       2017-01-04
4  2   2         2017-01-02     2017-01-03             <NA>
6  2   3         2017-01-03     2017-01-04       2017-01-05
9  3   4         2017-01-02     2017-01-03       2017-01-04