r - 以编程方式在 data.table 中使用 by 对列进行操作

Question

我想在 by 中使用包含列名的字符向量在 data.table以及定义组的交互方式。该向量包含在少数 data.table 中通用的列s，但每个 data.table有几个独特的列。那可能吗？下面举例。

library(data.table)
mtcarsdt <- data.table(mtcars)
bycols <- c('cyl', 'gear')   # Defined for use across multiple data.tables
mtcarsdt[
  , .(mpg = mean(mpg)),      # This does not work.
  by = c('carb%%2', bycols)  # How can I make this work?
]
mtcarsdt[
  , .(mpg = mean(mpg)), 
  by = .(carb%%2, cyl, gear) # This works
]

Answer 1

这是一种非常直接的方法:

mtcarsdt[, .(mpg = mean(mpg)), by = eval(as.call(parse(text = c(".", bycols, "carb %% 2"))))]
#    cyl gear carb      mpg
# 1:   6    4    0 19.75000
# 2:   4    4    1 29.10000
# 3:   6    3    1 19.75000
# 4:   8    3    0 14.63333
# 5:   4    4    0 24.75000
# 6:   8    3    1 16.30000
# 7:   4    3    1 21.50000
# 8:   4    5    0 28.20000
# 9:   8    5    0 15.40000
#10:   6    5    0 19.70000

另一种选择是构建整个表达式并对其进行评估/解析:

bycols = "cyl, gear"
eval(parse(text = paste0('mtcarsdt[, .(mpg = mean(mpg)), by = .(carb %% 2, ', bycols, ')]')))

您也可以使用 eval 玩相同的技巧。/ quote .

如果你不在乎保留 bycols列作为列，并且主要关心分组，您还可以执行以下操作:

mtcarsdt[, byvals := paste(.BY, collapse = ","), by = bycols][
         , .(mpg = mean(mpg)), by = .(byvals, carb %% 2)]
#    byvals carb      mpg
# 1:    6,4    0 19.75000
# 2:    4,4    1 29.10000
# 3:    6,3    1 19.75000
# 4:    8,3    0 14.63333
# 5:    4,4    0 24.75000
# 6:    8,3    1 16.30000
# 7:    4,3    1 21.50000
# 8:    4,5    0 28.20000
# 9:    8,5    0 15.40000
#10:    6,5    0 19.70000