按同一列聚合列

为什么传统的列聚合语法 data.table dt[, sum(x), by = "y"] 如果我们在 j 和 by ,即 dt[, sum(x), by = "x"] ?

library(data.table)

set.seed(1)
dt <- data.table(x = sample(c(1:10), 20, T), y = sample(letters[1:4], 20, T))
setorderv(dt, "y")

我想和 x 通过 X 但以下内容不起作用,它只是复制了 X 专栏:

> dt[, sum(x, na.rm = T), by = "x"]
     x V1
 1:  4  4
 2: 10 10
 3:  3  3
 4:  9  9
 5:  7  7
 6:  1  1
 7:  8  8
 8:  6  6
 9:  2  2
10:  5  5

如果我这样做:

> dt[, .(res = lapply(.SD, sum, na.rm = T)), by = 'x', .SDcols = "x"] 
     x res
 1:  4  12
 2: 10  30
 3:  3   9
 4:  9   9
 5:  7  21
 6:  1   1
 7:  8  24
 8:  6   6
 9:  2   2
10:  5   5

那是有效的。

另一方面,如果 通过 参数是一个不同于用于聚合的列 J :

> dt[, sum(x, na.rm = T), by = "y"]
   y V1
1: a 38
2: b 38
3: c 17
4: d 26