r语言 - 带ifelse和apply的二元操作符的非数值实参

Question

我希望没有人对我感到厌倦。下面我给出了一个完全虚构的问题的例子。只有在数据框架中包含数字列和字符串列的混合后才会出现错误，即使函数中排除了数字列(下面没有显示)。该函数在列c中查找d的内容，反之亦然，使用grepl，然后应该从ifelse中创建一个新的"yes"one_answers"no"语句列。我需要ifelse在测试后返回"是"one_answers"否"，但这是给出错误的部分。

a <- c(5:10)
b <- c(105:110)
c <- c("a","b","c","d","e","f")
d <- c("aa","bc","cd","ff","ee", "gf")
df <- data.frame(a,b,c,d)
newfunction <- function(x, col1, col2, col3, col4){ifelse(((sapply(lapply(x[[col4]], grepl, 
x[[col3]]),any)) | (sapply(lapply(x[[col3]], grepl, x[[col4]]),any))), (11 - x[[col1]]), (1 - 
x[[col2]]))}
df$new <- apply(df, 1, newfunction, "a", "b", "c", "d")

Error in 11 - x[[col1]] : non-numeric argument to binary operator

Answer 1

进入问题。解决这个问题的最佳方法是使用包dplyr中的函数case_when。我使用了stringr中的str_detect函数。

> library(tidyverse)
> library(stringr)
> 
> a  <- c(5:10)
> b  <- c(105:110)
> cc <- c("a","b","c","d","e","f")
> d  <- c("aa","bc","cd","ff","ee", "gf")
> df <- data.frame(a, b, cc, d)
> 
> 
> 
> df %>% mutate(case_when((str_detect(cc, d) | str_detect(d, cc)) ~ 'Yes', 
+                         TRUE ~ 'No'))
a   b cc  d case_when(...)
1  5 105  a aa            Yes
2  6 106  b bc            Yes
3  7 107  c cd            Yes
4  8 108  d ff             No
5  9 109  e ee            Yes
6 10 110  f gf            Yes

这个错误是由于您使用apply而导致的，它强制矩阵为字符矩阵(将数值转换为字符)。例如,

> apply(df, 2, function(x) x)
a    b     cc  d   
[1,] " 5" "105" "a" "aa"
[2,] " 6" "106" "b" "bc"
[3,] " 7" "107" "c" "cd"
[4,] " 8" "108" "d" "ff"
[5,] " 9" "109" "e" "ee"
[6,] "10" "110" "f" "gf"

一些指针，因为你似乎是新的r。首先，不要使用c作为一个名称，它是非常常用的函数来组合元素到一个向量。其次，函数的编写方式很难读懂。你应该把它分成几个步骤，这样更容易。

Answer 2

您还可以向量化grepl以简化代码:

Vgrepl <- Vectorize(grepl)
TF <- Vgrepl(pattern=df$c, x=df$d) | Vgrepl(pattern=df$d, x=df$c)
df$comp <- ifelse(TF, "Yes", "No")
df
#    a   b c  d comp
# 1  5 105 a aa  Yes
# 2  6 106 b bc  Yes
# 3  7 107 c cd  Yes
# 4  8 108 d ff   No
# 5  9 109 e ee  Yes
# 6 10 110 f gf  Yes