什么是允许参数通过...通过具有 purrr 映射的多个函数的正确方法

Question

考虑以下三个函数：

f1 <- function(df, ...) {
  df %>%
    mutate(
      model = map(.$splits, ~ f2(.))
    )
}

f2 <- function(split_df) {
  split_df %>%
    analysis() %>% 
    nest() %>%
    mutate(
      forecast = map(data, ~ f3(.$actuals))
    )
}

f3 <- function(x, N = 3, fn = mean, window_size = 6, ...) {
  # stuff
}

我在更大的管道中运行这些函数，但我的问题的要点是：我想将参数传递给f3()（例如，在以下代码行中将fn = mean 更改为fn = median：

df_resamples %>%
  mutate(
    result = .$resamples %>% map(~ f1(.))
  )

但是，以下都没有返回正确的结果：

# INCORRECT
df_resamples %>%
  mutate(
    result = .$resamples %>% map(~ f1(.), fn = median)
  )
# ALSO INCORRECT
df_resamples %>%
  mutate(
    result = .$resamples %>% map(~ f1(., fn = median))
  )

如果我在f2 中硬编码fn，我会得到正确的结果：

f2_good <- function(split_df) {
  split_df %>%
    analysis() %>% 
    nest() %>%
    mutate(
      forecast = map(data, ~ f3(.$actuals, fn = median))
    )
}

如果我从f1 传递...，为什么我不能传递fn = median 到f3？

注意：我能够解决将参数传递给 f1 和 f2 的问题，但这感觉不是很干，我想我可以使用省略号来规避这种方法：

f1 <- function(df, N = 3, fn = mean, window_size = 6) {
  df %>%
    mutate(
      model = map(.$splits, ~ f2(., N = N, fn = fn, window_size = window_size))
    )
}

f2 <- function(split_df, N = 3, fn = mean, window_size = 6) {
  split_df %>%
    analysis() %>% 
    nest() %>%
    mutate(
      forecast = map(data, ~ f3(.$actuals, N = N, fn = fn, window_size = window_size))
    )
}

做到以上几点，我可以做到：

df_resamples %>%
  mutate(
    result = .$resamples %>% map(~ f1(., fn = median))
  )

并获得预期的结果，但我必须将默认值传递给 f1 和 f2...

Answer 1

您是否尝试将... 添加到每个签名和函数调用中，例如：

f1 <- function(df, ...) {
  df %>%
    mutate(
      model = map(.$splits, ~ f2(., ...))
    )
}

f2 <- function(split_df, ...) {
  split_df %>%
    analysis() %>% 
    nest() %>%
    mutate(
      forecast = map(data, ~ f3(.$actuals, ...))
    )
}

f3 <- function(x, N = 3, fn = mean, window_size = 6, ...) {
  # stuff
}

然后调用

df_resamples %>%
  mutate(
    result = .$resamples %>% map(~ f1(., fn = median))
  )

编辑：玩具示例

f1 <- function(x, ...) f2(x, ...)
f2 <- function(x, ...) f3(x, ...)
f3 <- function(x, fn = mean, na.rm = FALSE, ...) fn(x, na.rm = na.rm) 

f1(c(1,2,3,NA), fn = max, na.rm = TRUE)