基于列表对象的子集数据框

Question

我在Turn 列中有一个带有语音数据的数据框：

test <- data.frame(
  Turn = c("Hi. I'm you an' you are me cos",
          "she'd've been so happy cos with all this stuff goin' on",
          "but we're in this together, because y' know things happens",
          "so you can't, cos well, ah because you know why!",
          "not now because it's too late!"), stringsAsFactors = F)

我想对那些在cos 和/或because 之前至少有四个字的行上的数据框进行子集化。为此，我在Turn 中计算了cos 和because 的索引：

test$Index <- sapply(strsplit(test$Turn, " "), function(x) which(x == 'cos'|x == 'because'))
test
                                                        Turn Index
1                             Hi. I'm you an' you are me cos     8
2    she'd've been so happy cos with all this stuff goin' on     5
3 but we're in this together, because y' know things happens     6
4           so you can't, cos well, ah because you know why!  4, 7
5                             not now because it's too late!     3

一行中有多个索引。这就是为什么我尝试这样的子集化失败的原因：

test[test$Index >= 5,]
Error in `[.data.frame`(test, test$Index >= 5, ) : 
  (list) object cannot be coerced to type 'double'

如何通过忽略第二个列出的Index 值来对test 进行子集化？

预期结果：

test
                                                        Turn Index
1                             Hi. I'm you an' you are me cos     8
2    she'd've been so happy cos with all this stuff goin' on     5
3 but we're in this together, because y' know things happens     6

我将不胜感激任何答案，包括不通过索引使用绕行但在子集过程中使用regex 模式的答案。

编辑：

sapply 范式中的解决方案非常简单，只需选择所列对象的第一个值：

sapply(test$Index, function(x) x[1])
[1] 4 5 6 4 3

Answer 1

我希望这会给你一个想法：

test <- data.frame(
  Turn = c("Hi. I'm you an' you are me cos",
          "she'd've been so happy cos with all this stuff goin' on",
          "but we're in this together, because y' know things happens",
          "so you can't, cos well, ah because you know why!",
          "not now because it's too late!"), stringsAsFactors = F)
rx <- "^\\s*(?:\\S+\\s+){0,3}(?:cos|because)\\b.*(*SKIP)(*F)|(?:\\S+[\\s,]+){4}\\b(cos|because)\\b"
Turn <- test[grepl(rx, test$Turn, perl=TRUE),]
split <- strsplit(Turn, "\\b(cos|because)\\b")
Index <- sapply(split, function(x) lengths(strsplit(trimws(x[[1]]), "\\s+"))+1)
test <- data.frame(Turn, Index, stringsAsFactors = F)
test

输出：

                                                       Turn Index
1                             Hi. I'm you an' you are me cos     8
2    she'd've been so happy cos with all this stuff goin' on     5
3 but we're in this together, because y' know things happens     6

请参阅R demo 和main regex demo。

正则表达式详细信息：

• ^\s*(?:\S+\s+){0,3}(?:cos|because)\b.*(*SKIP)(*F) - 匹配stirng 的开头，然后是零到三个单词，然后是cos 或because 作为整个单词和字符串的其余部分，然后跳过匹配项
• | - 或
• (?:\S+[\s,]+){4}\b(cos|because)\b - 匹配 cos 或 because 前面有四个单词。

Answer 2

基于 tidyverse 的解决方案如下所示。

library(dplyr)
library(purrr)
library(stringr)

test %>%
  mutate(index = map(str_split(Turn, ' '), 
                     ~ str_which(., 'cos|because')[1])) %>%
  filter(index >= 5)

#                                                         Turn index
# 1                             Hi. I'm you an' you are me cos     8
# 2    she'd've been so happy cos with all this stuff goin' on     5
# 3 but we're in this together, because y' know things happens     6