将`on`参数与CJ结合使用

Question

我想为多个不同长度的变量的子集导出摘要统计。 尝试使用on中的data.table参数与CJ结合使用，我得到了一个解决方案，其打字工作比使用键控变量要多得多。有没有更好的解决方案来完成这项任务？

require(data.table) 
dt <- CJ(year = 2015:2016, class = 1:4, age = 15:20)
set.seed(1)
dt[, var := rnorm(48)]
dt[CJ(class = 2:3, age = 15:17),
      list(med = median(var)), on = c("class", "age"),
      keyby = .(age, year, class)]

##     age year class        med
##  1:  15 2015     2  0.4874291
##  2:  15 2015     3 -0.6212406
##  3:  15 2016     2  1.3586796
##  4:  15 2016     3 -0.3942900
##  5:  16 2015     2  0.7383247
##  6:  16 2015     3 -2.2146999
##  7:  16 2016     2 -0.1027877
##  8:  16 2016     3 -0.0593134
##  9:  17 2015     2  0.5757814
## 10:  17 2015     3  1.1249309
## 11:  17 2016     2  0.3876716
## 12:  17 2016     3  1.1000254

##### keyed version -- less typing work
setkey(dt, class, age)
dt[CJ(2:3, 15:17), list(med = median(var)),
   keyby = .(age, year, class)]

##### same result as above