使用不纯函数循环遍历数据帧行的最优雅方法是什么？

Question

如果我有下面这段代码：

my_func <- function(var1, var2, var3, var4) {
    ... (side effect included) 
}

df <- crossing(
    nesting(var1=...,var2=....)
    nesting(var3=...,var4=....)
)

将 my_func 应用于 df 的每一行的最优雅的方法是什么？ plus my_func 不是一个纯函数，它被设计来进行一些副作用（IO, plot ...）

方法一

my_func_wrapper <- function(row) {
  my_func(row['var1'], row['var2'], row['var3'], row['var4'])
}

# Vector coercion is a problem, if variables are not the same type.
apply(df, 1, my_func_wrapper)

方法二

df %>%
  rowwise() %>%
  do(result=invoke(my_func, .)) %>% #If it ends here, I will be pretty happy.
  .$result # Relying auto print feature to plot or trigger some side effect

方法三

#This looks pretty good on its own but it does not play well with the pipe %>%
foreach(row=iter(df, by='row'))  %do% invoke(my_func, row)

#Method 3.1 (With Pipe)
 df %>%
   (function(df) foreach(row=iter(df, by='row'))  %do% invoke(my_func, row))

#Method 3.2 this does not work
# df %>%
#   foreach(row=iter(., by='row'))  %do% invoke(my_func, row)

#Method 3.3 this does not work
#I am trying to get this work with purrr's simplified anonymous function, but it does not work.
# df %>%
#    as_function(~ foreach(row=iter(., by='row'))  %do% invoke(my_func, row))

有没有更好的方法，可以和 %>% 一起玩？

Answer 1

老实说，我会用 purr 的 pmap::pmap

library(tidyverse)

df = data.frame(
  x = rnorm(10),
  y = runif(10)
)
df %>% 
  pmap_dbl(function(x, y) {
    min(x,y)
  })

Answer 2

我发现 tidyverse 提供的许多此类操作仍然比 plyr 差。示例：

> library(plyr)
> library(tidyverse)
> #dummy function
> your_function = function(...) {do.call(args = list(..., sep = " + "), what = str_c)}
> alply(mpg[1:5, ], .margins = 1, .fun = function(row) {
+   your_function(row$manufacturer, row$cyl, row$trans)
+ }) %>% unlist()
                      1                       2                       3                       4                       5 
  "audi + 4 + auto(l5)" "audi + 4 + manual(m5)" "audi + 4 + manual(m6)"   "audi + 4 + auto(av)"   "audi + 6 + auto(l5)"

我不知道你想从函数中收集什么，但你可能想要 alply() 或 adply()。

使用不纯函数循环遍历数据帧行的最优雅方法是什么？

What is most elegant way to loop through rows of a data frame with an impure function?

foreach

r

dplyr

tidyverse