在R函数中无需显式引用数据框直接调用列的更佳实现方法

我希望在R函数里直接通过列名调用数据框的列（仅写列名即可），而非每次都写df$columnname或df["columnname"]。使用attach()函数可实现此需求，但听说该函数存在性能不佳等问题。目前我实现的代码及调用输出如下：

第一次调用代码与输出 #

> somefunction <- function(df1,df2) {attach(df1); attach(df2); mean(Sepal.Length) + speed*3}
> somefunction(iris, cars)
The following objects are masked from iris:

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from b (pos = 5):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from a (pos = 6):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from a (pos = 8):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from cars:

    dist, speed

The following objects are masked from b (pos = 8):

    dist, speed

 [1] 17.84333 17.84333 26.84333 26.84333 29.84333 32.84333 35.84333 35.84333 35.84333 38.84333 38.84333 41.84333 41.84333 41.84333 41.84333 44.84333
[17] 44.84333 44.84333 44.84333 47.84333 47.84333 47.84333 47.84333 50.84333 50.84333 50.84333 53.84333 53.84333 56.84333 56.84333 56.84333 59.84333
[33] 59.84333 59.84333 59.84333 62.84333 62.84333 62.84333 65.84333 65.84333 65.84333 65.84333 65.84333 71.84333 74.84333 77.84333 77.84333 77.84333
[49] 77.84333 80.84333

再次调用的输出 #

> somefunction(iris, cars)
The following objects are masked from df1 (pos = 4):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from iris:

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from b (pos = 7):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from a (pos = 8):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from a (pos = 10):

    Petal.Length, Petal.Width, Sepal.Length, Sepal.Width, Species

The following objects are masked from df2 (pos = 4):

    dist, speed

The following objects are masked from cars:

    dist, speed

The following objects are masked from b (pos = 10):

    dist, speed

 [1] 17.84333 17.84333 26.84333 26.84333 29.84333 32.84333 35.84333 35.84333 35.84333 38.84333 38.84333 41.84333 41.84333 41.84333 41.84333 44.84333
[17] 44.84333 44.84333 44.84333 47.84333 47.84333 47.84333 47.84333 50.84333 50.84333 50.84333 53.84333 53.84333 56.84333 56.84333 56.84333 59.84333
[33] 59.84333 59.84333 59.84333 62.84333 62.84333 62.84333 65.84333 65.84333 65.84333 65.84333 65.84333 71.84333 74.84333 77.84333 77.84333 77.84333
[49] 77.84333 80.84333

可以看到调用时出现大量变量遮蔽警告，且pos值会随调用次数变化，并非理想情况。请问是否有更好的实现方法？

可行替代方案 #

1. 使用with()函数 #

with()可以指定数据框作为临时环境，在括号内直接引用列名，不会污染全局环境，也不会出现变量遮蔽问题：

somefunction <- function(df1, df2) {
  # 从df1中计算Sepal.Length的均值
  mean_sepal <- with(df1, mean(Sepal.Length))
  # 在df2的环境中计算结果
  with(df2, mean_sepal + speed * 3)
}

somefunction(iris, cars)

2. 使用tidyverse系列工具（dplyr） #

现代R常用的tidyverse语法支持直接引用列名，代码可读性高，适合数据处理场景：

library(dplyr)

somefunction <- function(df1, df2) {
  mean_sepal <- df1 %>% pull(Sepal.Length) %>% mean()
  df2 %>% transmute(result = mean_sepal + speed * 3) %>% pull()
}

somefunction(iris, cars)

3. 使用data.table #

data.table不仅支持直接引用列名，还具备高效的大数据处理能力：

library(data.table)

somefunction <- function(df1, df2) {
  dt1 <- as.data.table(df1)
  dt2 <- as.data.table(df2)
  
  mean_sepal <- dt1[, mean(Sepal.Length)]
  dt2[, mean_sepal + speed * 3]
}

somefunction(iris, cars)

4. 自定义临时环境（进阶） #

如果需要模拟attach()的全局式引用，可创建临时环境存放数据框列，用完即销毁，避免污染全局环境：

somefunction <- function(df1, df2) {
  temp_env <- new.env()
  # 将两个数据框的列导入临时环境
  list2env(as.list(df1), envir = temp_env)
  list2env(as.list(df2), envir = temp_env)
  
  # 在临时环境中计算结果
  result <- with(temp_env, mean(Sepal.Length) + speed * 3)
  
  # 销毁临时环境
  rm(temp_env)
  result
}

somefunction(iris, cars)

内容的提问来源于stack exchange，提问作者Aku-Ville Lehtimäki