在以两列为条件的行序列上过滤数据框答案

【问题标题】：Filter dataframe on sequence of rows conditional on two columns在以两列为条件的行序列上过滤数据框
【发布时间】：2023-04-11 12:32:01
【问题描述】：

我有这种类型的数据，其中Sequ 列中的数值定义了一个行序列，Q 中的字符值命名了序列的类型：

df <- data.frame(
  Line = 1:12,
  Speaker = c(NA, "ID01.A", NA, "ID01.B", "ID07.A", NA, "ID33.B", 
              "ID33.A", "ID33.C", NA, "ID77.A", "ID77.C"),
  Utterance = c(NA, "Who did it?", "(1.99)", "Peter did.", "Hello!", NA, "So you're coming?", 
                "erm", "Yes, sure.", "(0.22)", "Good night?", "Yeah, sleep well"),
  Sequ = c(NA,1,1,1, NA,NA, 2,2,2, NA, 3,3),
  Q = c(NA, "q_wh", "", "", NA, NA, "q_decl", "", "", NA, "q_wh", "")
)

我想对那些 Sequ 数值（而不是 NA）和其中 Q == q_wh 的数据框进行子集化。我可以使用na_if 然后fill 来完成这项任务：

library(tidyr)
df %>%
  mutate(Q = na_if(Q, "")) %>%
  fill(Q, .direction = "down") %>%
  filter(!is.na(Sequ) & Q == "q_wh")
  Line Speaker        Utterance Sequ    Q
1    2  ID01.A      Who did it?    1 q_wh
2    3    <NA>           (1.99)    1 q_wh
3    4  ID01.B       Peter did.    1 q_wh
4   11  ID77.A      Good night?    3 q_wh
5   12  ID77.C Yeah, sleep well    3 q_wh

但是有没有另一种更直接的方法，不绕道na_if和fill，过滤df？

编辑：

我找到了一个没有fill 和na_if 的解决方案：

df %>%
  group_by(Sequ) %>%
  mutate(Q = Q[!Q==""]) %>%
  filter(!is.na(Sequ) & Q == "q_wh")

【问题讨论】：

标签： r tidyr

【解决方案1】：

使用ave 将"Sequ" 替换为第一个相应的Q，最后是subset。

df[!is.na(df$Sequ), ] |>
  transform(Q=ave(Q, Sequ, FUN=\(x) x[1])) |>
  subset(!is.na(Sequ) & Q == 'q_wh')
#    Line Speaker        Utterance Sequ    Q
# 2     2  ID01.A      Who did it?    1 q_wh
# 3     3    <NA>           (1.99)    1 q_wh
# 4     4  ID01.B       Peter did.    1 q_wh
# 11   11  ID77.A      Good night?    3 q_wh
# 12   12  ID77.C Yeah, sleep well    3 q_wh

注意：R version 4.1.2 (2021-11-01).

或者使用tidyverse方言。

library(magrittr)
df[!is.na(df$Sequ), ] %>%
  dplyr::mutate(Q=ave(Q, Sequ, FUN=\(x) x[1])) %>%
  dplyr::filter(!is.na(Sequ) & Q == 'q_wh')
#   Line Speaker        Utterance Sequ    Q
# 1    2  ID01.A      Who did it?    1 q_wh
# 2    3    <NA>           (1.99)    1 q_wh
# 3    4  ID01.B       Peter did.    1 q_wh
# 4   11  ID77.A      Good night?    3 q_wh
# 5   12  ID77.C Yeah, sleep well    3 q_wh

数据：

df <- structure(list(Line = 1:12, Speaker = c(NA, "ID01.A", NA, "ID01.B", 
"ID07.A", NA, "ID33.B", "ID33.A", "ID33.C", NA, "ID77.A", "ID77.C"
), Utterance = c(NA, "Who did it?", "(1.99)", "Peter did.", "Hello!", 
NA, "So you're coming?", "erm", "Yes, sure.", "(0.22)", "Good night?", 
"Yeah, sleep well"), Sequ = c(NA, 1, 1, 1, NA, NA, 2, 2, 2, NA, 
3, 3), Q = c(NA, "q_wh", "", "", NA, NA, "q_decl", "", "", NA, 
"q_wh", "")), class = "data.frame", row.names = c(NA, -12L))

【讨论】：

我有一个旧版本的 R。我猜|> 相当于 %>%，但什么相当于 `FUN=` 中的反斜杠？
@ChrisRuehlemann 是的，|> 是新的基础管道，将 \ 替换为 function 即可。不过，我建议您或您的管理员更新 R。请注意，我已经更新了 tidyverse 版本。
谢谢。我赞成该解决方案。然而问题是Q 列不能填充诸如q_wh 之类的值；在序列的开头必须有一个这样的值。（这也是我的“填充”解决方案不正确的原因。）
@ChrisRuehlemann 你能举一个我们的解决方案会失败的例子吗？
不，只是我需要在转换后对数据进行后处理，这样不填满所有Q行会方便得多。