如何向数据框列表批量应用带多参数的自定义函数？

问题场景 #

我有一个包含大量列标题完全一致的数据框的列表，希望对每个数据框应用同一个处理函数。尝试将单个数据框的处理代码封装成函数后，用purrr::map()应用到列表时，多次出现“Epoch或Event不存在”的错误，用.或~作为占位符也没用。

示例数据 #

df <- data.frame(Epoch = c(99,126,136,148,150,200,228,247,268,269,285,285,294,308,309,319,320,324,326,337,338,352,353,380,382,418,419,424,431,437,440,447,449,470,515,548,550,550,561,589,590,596,598,603,612,616,623,626,633,655,663,668,682,687,704,708,717,730,744,752,771,794,810,811,816,819,825,832,832,841,847,851,853,868,871,881,893,915,918,922,935,943),
                Event = c( "hypopnea","auto_trig","double_trig","auto_trig","double_trig"   , "double_trig" ,"hypopnea" ,"hypopnea","auto_trig" ,"double_trig","auto_trig","double_trig"  , "double_trig" ,"auto_trig", "double_trig","auto_trig", "double_trig","auto_trig", "double_trig" , "double_trig", "hypopnea","hypopnea" ,"double_trig","double_trig","hypopnea", "double_trig" ,"hypopnea" , "double_trig", "high_leak", "auto_trig", "double_trig", "auto_trig" ,"double_trig", "double_trig", "double_trig", "hypopnea", "auto_trig",  "double_trig",  "hypopnea",  "double_trig", "auto_trig",    "double_trig", "auto_trig","double_trig","double_trig","auto_trig", "auto_trig","hypopnea",    "double_trig", "double_trig", "double_trig", "auto_trig", "double_trig","auto_trig", "hypopnea" , "double_trig", "double_trig", "double_trig", "double_trig", "ineffective_eff", "ineffective_eff", "double_trig", "double_trig", "high_leak", "double_trig",  "hypopnea", "auto_trig", "hypopnea",        "double_trig", "high_leak", "double_trig", "high_leak", "auto_trig", "double_trig", "double_trig",   "double_trig", "double_trig", "auto_trig", "double_trig", "double_trig", "double_trig", "double_trig"))

df2 <- data.frame(Epoch = c(98,99,126,135,141,181,183,200,236,247,257,284,308,353,353,354,380,382,418,418,431,561,733,751,841,851),
                 Event = c("double_trig", "hypopnea", "auto_trig", "hypopnea", "hypopnea", "double_trig", "hypopnea", "double_trig", "hypopnea", "hypopnea", "hypopnea", "double_trig", "hypopnea", "double_trig", "hypopnea", "double_trig", "hypopnea", "high_leak", "hypopnea", "hypopnea", "ineffective_eff", "high_leak", "high_leak", "double_trig", "double_trig", "ineffective_eff"))

List1 <- list(df, df2)

单个数据框的处理逻辑 #

df3 <- data.frame(Section = seq(1, max(df$Epoch/10+1), by = 1),
                Epoch5 = seq(1, max(df$Epoch), by =10),
                Epoch51 = seq(11, max(df$Epoch)+10, by = 10))

sp <- split(df$Epoch, df$Event)
nms <- names(sp)
E <- c(df3$Epoch5, max(df3$Epoch51))

df3 <- df3 |>
  cbind(
    sapply(nms, \(i) {
      sec <- findInterval(sp[[i]], E)
      df3$Section %in% sec |> as.integer()
    })
  )

rm(sp, nms, E)

解决方案 #

错误原因 #

你封装函数时，大概率没有把代码中所有引用df的地方替换成函数的参数，导致函数在全局环境中查找df，而非使用map传入的每个数据框，因此触发“Epoch或Event不存在”的错误。

正确的函数封装与调用 #

将处理逻辑封装成接受单个数据框参数的函数，内部所有依赖原数据框的部分都改用参数引用：

library(purrr)

# 封装处理函数
process_df <- function(x) {
  # 创建初始分组数据框
  df3 <- data.frame(
    Section = seq(1, max(x$Epoch/10 + 1), by = 1),
    Epoch5 = seq(1, max(x$Epoch), by = 10),
    Epoch51 = seq(11, max(x$Epoch) + 10, by = 10)
  )
  
  # 按Event拆分Epoch
  sp <- split(x$Epoch, x$Event)
  nms <- names(sp)
  E <- c(df3$Epoch5, max(df3$Epoch51))
  
  # 绑定统计列
  df3 <- df3 |>
    cbind(
      sapply(nms, \(i) {
        sec <- findInterval(sp[[i]], E)
        as.integer(df3$Section %in% sec)
      })
    )
  
  # 清理临时变量（可选，函数内的变量会自动回收）
  rm(sp, nms, E)
  
  return(df3)
}

# 应用到数据框列表
result_list <- map(List1, process_df)

验证结果 #

运行上述代码后，result_list中的每个元素就是对应数据框处理后的结果，不会再出现找不到列的错误。

内容的提问来源于stack exchange，提问作者AR459