泛化用于自定义函数的for循环 [英] Generalize a for-loop for use in a custom function
问题描述
使用下面的for循环,我可以创建给定员工之上的所有经理的列表(基本上是员工经理,经理经理等的列表)
Using the for-loop below I can create a list of all managers above a given employee (essentially a list of an employee's manager, her manager's manager, etc.)
library(dplyr)
library(tidyr)
library(purrr)
# Create test data
ds <-
tibble(
emp_id = c("001", "002", "003", "004", "005"),
mgr_id = c("002", "004", "004", "005", NA)
)
# Hardcoded for-loop example
mgr_ids_above <- vector("list", length = 5)
id <- "001"
for (i in seq_along(mgr_ids_above)) {
mgr_ids_above[[i]] <- ds$mgr_id[ds$emp_id == id]
id <- mgr_ids_above[[i]]
}
# drop NAs
mgr_ids_above <- unlist(mgr_ids_above)
mgr_ids_above <- mgr_ids_above[!is.na(mgr_ids_above)]
# return to list format
as.list(mgr_ids_above)
我希望将此for循环应用于整个数据框,并将结果保存在列表列中.我可以使用pmap()
成功地执行此操作,以将硬编码的for循环应用于我的数据帧,但是当我尝试编写通用函数时,一切都崩溃了.
My hope is to apply this for-loop to the entire data frame and save the results in a list-column. I can successfully do this using pmap()
to apply a hard-coded for-loop to my data frame, but when I try to write a generalized function, everything falls apart.
# Define custom function with hardcoded data and variable names
get_mgrs_above <- function(id, max_steps = 5){
mgr_ids_above <- vector("list", length = max_steps)
for (i in seq_along(mgr_ids_above)) {
mgr_ids_above[[i]] <- ds$mgr_id[ds$emp_id == id]
id <- mgr_ids_above[[i]]
}
# drop NAs
mgr_ids_above <- unlist(mgr_ids_above)
mgr_ids_above <- mgr_ids_above[!is.na(mgr_ids_above)]
# return to list format
as.list(mgr_ids_above)
}
# Apply custom function
ds_mgrs_above <-
ds %>%
mutate(
ranks_above = pmap(
list(id = emp_id),
get_mgrs_above
)
)
以上代码的输出为
A tibble: 5 x 3
emp_id mgr_id ranks_above
<chr> <chr> <list>
1 001 002 <list [3]>
2 002 004 <list [2]>
3 003 004 <list [2]>
4 004 005 <list [1]>
5 005 NA <list [0]>
ranks_above
列表列的内容如下
ds_mgrs_above$ranks_above[[1]]
[[1]]
[1] "002"
[[2]]
[1] "004"
[[3]]
[1] "005"
我将所有数据和变量作为参数提供的失败函数失败,并显示消息"mutate_impl(.data,点)中的错误: 评估错误:元素1的长度为2,而不是1或5.:
My failing function with all data and variables supplied as arguments fails with the message, "Error in mutate_impl(.data, dots) : Evaluation error: Element 1 has length 2, not 1 or 5..":
get_mgrs_above <- function(
data,
id = emp_id,
mgr_id = mgr_id,
emp_id = emp_id,
max_steps = 5){
mgr_ids_above <- vector("list", length = max_steps)
for (i in seq_along(mgr_ids_above)) {
mgr_ids_above[[i]] <- data$mgr_id[data$emp_id == id]
id <- mgr_ids_above[[i]]
}
# drop NAs
mgr_ids_above <- unlist(mgr_ids_above)
mgr_ids_above <- mgr_ids_above[!is.na(mgr_ids_above)]
# return to list format
as.list(mgr_ids_above)
}
ds %>%
mutate(
ranks_above = pmap(
list(
data = ds,
id = emp_id,
mgr_id = mgr_id,
emp_id = emp_id,
max_steps = 5
),
get_mgrs_above
)
)
为避免混淆,这是一篇有关如何编写可归纳函数的文章,该函数将从两列创建一个列表列.这是对拥有约1.5万名员工的数据帧进行较大数据整理尝试的一个组成部分.
To avoid confusion, this is a post about how to write a generalizable function that will create a list column from two columns. This is one component of a larger data munging attempt on a data frame with ~15k employees.
推荐答案
循环的构造方式使我感到困惑,因此我使用while
进行了更改.最后只允许使用map
.
The way your loop was structured was confusing me so I changed it by using while
. This allows just an application of map
at the end.
get_mgrs_above <- function(id, data = NULL, max_steps = 5) {
stopifnot(!is.null(data))
mgr_ids_above <- list()
mgr <- id
iter <- 0
while (iter < max_steps & !is.na(mgr)) {
mgr <- data$mgr_id[data$emp_id == mgr]
if (!is.na(mgr)) {
mgr_ids_above <- append(mgr_ids_above, mgr)
}
iter <- iter + 1
}
return(mgr_ids_above)
}
ds$ranks_above <- map(ds$emp_id, get_mgrs_above, data = ds)
这篇关于泛化用于自定义函数的for循环的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!