R语言scout函数报错排查：condition has length >1及NA警告

问题背景 #

我有一份涵盖五大联赛所有足球球员的数据集，正在构建scout函数，用来筛选出所选指标处于前85百分位的球员名单。测试调用scout(Total_Big_5_new,"Nutmegs")时出现报错：

the condition has length > 1
In addition: Warning message:
In percentile(database$metric) : NAs introduced by coercion

现有代码 #

scout函数 #

scout <- function(database, ...) {
  l <- list(...)
  l2 <- list()
  j <- 1
  for(metric in l){
    if(metric %in% colnames(database)){
      l2[[j]] <-  percentile(database[[metric]])
      j <- j + 1
    }else{
      print(paste("The stat", metric, "is not recorded"))
    }
  }
  i <- 1
  k <- 1
  shortlist <- list()
  for (player in database){
    compared <- select(database, unlist(l))
    if (all(compared) > all(unlist(l2))){
      shortlist[[i]] <- player
      i <- i + 1
    }
  }
  return(shortlist)
  }

percentile函数 #

percentile <- function(metric, value = 0.85) {
   answer <- unname(quantile(metric, c(value)))
   return(as.numeric(paste(answer)))
 }

示例数据集与预期结果 #

构造随机数据集df：

df <- as_tibble(data.frame(
  Player  = c(LETTERS[1:13]),
  Goals = c(sample(1:45, 13, replace=FALSE)),
  Assists = c(sample(1:31, 13, replace=FALSE)),
  Nutmegs = c(sample(1:28, 13, replace = FALSE)),
  Dribbles = c(sample(43:208, 13, replace = FALSE))
))

数据集内容：

Player Goals Assists Nutmegs Dribbles
   <chr>  <int>   <int>   <int>    <int>
 1 A         23      16       1      125
 2 B          7       2      19      195
 3 C         21       4      28      142
 4 D         28      19      23      112
 5 E          8      27      26      152
 6 F         17      23      16       45
 7 G         30       6      25      206
 8 H         26      24       8      136
 9 I         18       3      27       99
10 J         31      25       7      198
11 K          4      21      13       82
12 L          1      13      22       66
13 M         43       7       4      194

调用percentile(df$Goals, 0.65)返回25.4，预期调用scout(df,"Goals")应返回球员D、G、H、J、M。

报错原因分析 #

1. 遍历逻辑错误：for (player in database)会遍历数据集的列而非行，导致后续筛选逻辑完全偏离预期。
2. 条件判断逻辑错误：all(compared) > all(unlist(l2))会把整列数据转为单一逻辑值，无法实现逐行比较每个球员的指标是否超过对应百分位阈值。
3. 冗余数值转换：as.numeric(paste(answer))属于多余操作，quantile返回的数值可直接使用，该操作是NA警告的潜在来源。

修复后的代码 #

修复percentile函数 #

percentile <- function(metric, value = 0.85) {
  # 直接返回百分位结果，添加NA值处理
  unname(quantile(metric, value, na.rm = TRUE))
}

修复scout函数 #

library(dplyr)

scout <- function(database, ...) {
  metrics <- unlist(list(...))
  
  # 校验指标有效性
  valid_metrics <- metrics[metrics %in% colnames(database)]
  invalid_metrics <- setdiff(metrics, valid_metrics)
  if(length(invalid_metrics) > 0){
    lapply(invalid_metrics, function(m) {
      message(paste("The stat", m, "is not recorded"))
    })
  }
  
  if(length(valid_metrics) == 0) {
    message("No valid metrics provided")
    return(data.frame())
  }
  
  # 计算每个指标的85百分位阈值
  thresholds <- sapply(valid_metrics, function(m) {
    percentile(database[[m]])
  })
  
  # 筛选所有指标达标球员
  shortlist <- database %>%
    filter(across(all_of(valid_metrics), ~ .x >= thresholds[cur_column()]))
  
  return(shortlist)
}

验证结果 #

调用scout(df, "Goals")，会自动计算Goals的85百分位阈值，筛选出所有进球数≥该阈值的球员，符合预期结果。

内容的提问来源于stack exchange，提问作者med0504