data.table：分组依据，然后使用返回多个新列的自定义函数进行聚合

Question

在 data.table 中，我该如何执行以下操作：

组合 table 几列
然后将每个组交给自定义聚合函数，即：
从组 table 子集中获取所有列，并通过返回几个将添加到 table

这里的技巧是在不多次调用聚合函数的情况下生成几个新列。

示例：

library(data.table)
mtcars_dt <- data.table(mtcars)

returnsOneColumn <- function(dt_group_all_columns){
  "returned_value_1"
}

# works great, returns one new column as summary per group
mtcars_dt[,
          list( new_column_1 = returnsOneColumn(dt_group_all_columns= .SD) ),
          by = c("mpg", "cyl"),
          .SDcols = colnames(mtcars_dt)
          ]

returnsMultipleColumns <- function (dt_group_all_columns){
  list( "new_column_1" = "returned_value_1", 
        "new_column_2" = "returned_value_2"  )
}

# does not work: Ideally, I would like to have mpg, cyl, and several columns 
# generated from once calling returnsMultipleColumns
mtcars_dt[,
          list( returnsMultipleColumns(dt_group_all_columns = .SD) ),
          by = c("mpg", "cyl"),
          .SDcols = colnames(mtcars_dt)
          ]

# desired output should look like this
#
#     mpg cyl     new_column_1     new_column_2
# 1: 21.0   6 returned_value_1 returned_value_2
# 2: 22.8   4 returned_value_1 returned_value_2
# 3: 21.4   6 returned_value_1 returned_value_2
# 4: 18.7   8 returned_value_1 returned_value_2

相关：

Assign multiple columns using := in data.table, by group

Answer 1

您已经从函数中返回了一个列表。您无需再次列出它们。所以删除 list 并使用如下代码

mtcars_dt[,
           returnsMultipleColumns(dt_group_all_columns = .SD),
           by = c("mpg", "cyl"),
           .SDcols = colnames(mtcars_dt)
           ]
     mpg cyl     new_column_1     new_column_2
 1: 21.0   6 returned_value_1 returned_value_2
 2: 22.8   4 returned_value_1 returned_value_2
 3: 21.4   6 returned_value_1 returned_value_2
 4: 18.7   8 returned_value_1 returned_value_2

data.table：分组依据，然后使用返回多个新列的自定义函数进行聚合

data.table: Group by, then aggregate with custom function returning several new columns

grouping

aggregate

r

summary

data.table