基于字段特定值选择列：dplyr select(where)报错排查

我正在基于调查数据创建表格，该调查要求受访者对活动各元素的满意度进行评分。若受访者对某活动元素表示不满意，需填写自由文本子问题说明不满原因。

目前我已将所有活动元素及其评分格式化为单独列，同时保留了通用的自由文本不满反馈字段，并过滤掉了该字段中的NA值：

# 当前输出 

Response_ID | Reasons_for_Dissatisfaction | Rank_Food             | Rank_Facilities       | Rank_Content
----------------------------------------------------------------------------------------------------------     
1           | <Free-text feedback>        | Somewhat dissatisfied | Very dissatisfied     | Very satisfied 
----------------------------------------------------------------------------------------------------------
2           | <Free-text feedback>        | Very dissatisfied     | Somewhat dissatisfied | Very satisfied

我希望表格仅显示受访者“Somewhat dissatisfied”（有点不满意）或“Very dissatisfied”（非常不满意）的活动元素列，同时保留自由文本形式的评分原因列，目标输出如下：

# 目标输出

Response_ID | Reasons_for_Dissatisfaction | Rank_Food             | Rank_Facilities       
-----------------------------------------------------------------------------------------
1           | <Free-text feedback>        | Somewhat dissatisfied | Very dissatisfied
-----------------------------------------------------------------------------------------
2           | <Free-text feedback>        | Very dissatisfied     | Somewhat dissatisfied

我需要代码能随数据集更新自动调整：若新增反馈中出现新的不满意活动元素，需自动将该元素列及对应记录加入表格，因此不能使用绝对列引用。我尝试用select(where())筛选仅包含“Somewhat dissatisfied”或“Very dissatisfied”值的列，同时保留“Response_ID”和“Reasons_for_Dissatisfaction”列，但运行报错：

# 示例代码

satisfactionrank <- data.frame(
  `Response_ID`=c(1,2),
  `Reasons_for_Dissatisfaction`=c("<Free-text feedback>","<Free-text feedback>"),
  `Rank_Food`=c("Somewhat dissatisfied","Very dissatisfied"),
  `Rank_Facilities`=c("Very dissatisfied","Somewhat dissatisfied"),
  `Rank_Content`=c("Very satisfied","Very satisfied")
)

eventdissatisfaction <- satisfactionrank %>%
  select(
    `Response_ID`,
    `Reasons_for_Dissatisfaction`,
    where(contains("Rank") & any(. == "Somewhat dissatisfied") | any(. == "Very dissatisfied"))
  )

报错信息：

Error in `select()`:
ℹ In argument: `where(...)`.
Caused by error in `where()`:
! Can't convert `fn`, a logical vector, to a function.
Run `rlang::last_trace()` to see where the error occurred.

请问我的代码哪里出错了？

错误原因 #

你的代码存在两个核心问题：

1. where()使用逻辑错误：where()要求传入一个返回逻辑值的函数（或匿名函数）来判断列是否符合条件，但你直接将列名筛选函数contains("Rank")与列值判断逻辑混用，导致返回的是逻辑向量而非函数，触发类型转换错误。
2. 逻辑运算符优先级问题：&的优先级高于|，原表达式会被错误解析，无法正确匹配你需要的条件组合。

修正后的代码 #

library(dplyr)

satisfactionrank <- data.frame(
  `Response_ID`=c(1,2),
  `Reasons_for_Dissatisfaction`=c("<Free-text feedback>","<Free-text feedback>"),
  `Rank_Food`=c("Somewhat dissatisfied","Very dissatisfied"),
  `Rank_Facilities`=c("Very dissatisfied","Somewhat dissatisfied"),
  `Rank_Content`=c("Very satisfied","Very satisfied")
)

eventdissatisfaction <- satisfactionrank %>%
  select(
    Response_ID,
    Reasons_for_Dissatisfaction,
    # 先匹配列名含"Rank"的列，再判断该列是否存在不满意评分
    where(~ grepl("Rank", colnames(.)) & any(. %in% c("Somewhat dissatisfied", "Very dissatisfied")))
  )

print(eventdissatisfaction)

代码说明 #

1. where(~ ...)：用波浪线定义匿名函数，.指代当前列的所有值，colnames(.)指代当前列的列名。
2. grepl("Rank", colnames(.))：精准筛选所有活动元素评分列（列名含"Rank"）。
3. any(. %in% c(...))：判断当前列中是否存在任意一个"有点不满意"或"非常不满意"的评分，存在则保留该列。
4. 逻辑分组：通过括号明确条件执行顺序，确保先筛选列名，再判断列值，同时用%in%简化多值匹配的写法。

输出结果 #

Response_ID Reasons_for_Dissatisfaction             Rank_Food       Rank_Facilities
1           1          <Free-text feedback> Somewhat dissatisfied    Very dissatisfied
2           2          <Free-text feedback>    Very dissatisfied Somewhat dissatisfied

该代码可自动适配新增的Rank_*列：只要新列中存在不满意评分，就会被自动加入结果表格。

内容的提问来源于stack exchange，提问作者Mary Rachel