如何使用扩展/枢轴或重塑来变换分层数据帧

Question

我目前正在尝试以降序的分层格式重新格式化文档库。 docFrom列包含较高级别的文档，subDoc包含较低级别的文档，其父列表示该文档向下的级别数，其中1表示顶部的文档。数据Docs都是字符串，当前看起来像这样，唯一的区别是subDoc包含虚拟数据不会显示的所有唯一字符串，将它们想象为网络，节目和剧集的实际名称。

    docFrom      subDoc           Parent  
    NA           Network 1        1 
    Network 1    TvShow 1         2
    Network 1    TvShow 2         2             
    Network 1    TvShow 3         2
    Network 1    TvShow 4         2 
    TvShow 1     Episode 1        3 
    TvShow 1     Episode 2        3 
    TvShow 2     Episode 1        3 
    TvShow 2     Episode 2        3 
    TvShow 3     Episode 1        3 
    TvShow 1     Episode 2        3 

出于可视化目的，我想将其转换为

  1          2         3 
  Network 1  TvShow 1  Episode 1 
  Network 1  TvShow 1  Episode 2
  Network 1  TvShow 2  Episode 1 
  Network 1  TvShow 2  Episode 2 
  Network 1  TvShow 3  Episode 1 
  Network 1  TvShow 3  Episode 2 

使用df <- reshape(Docs,idvar = "docFrom", timevar = "Parent", direction = "wide")无效，

df <- spread(Docs, Parent, subDoc)

我试图找到解决方案，但找不到反映这种情况的任何数据。有什么功能可以用来重塑这样的数据框吗？

Answer 1

我们将结合使用基R和sqldf()包来解决此问题。我们可以使用Parent列将数据分为3个数据帧，并合并两个结果数据帧，其中Parent的电视节目名称为2或3。

textFile <- "docFrom  |subDoc   |Parent  
NA       |Network 1|1 
Network 1|TvShow 1 |2
Network 1|TvShow 2 |2             
Network 1|TvShow 3 |2
Network 1|TvShow 4 |2 
TvShow 1 |Episode 1|3 
TvShow 1 |Episode 2|3 
TvShow 2 |Episode 1|3 
TvShow 2 |Episode 2|3 
TvShow 3 |Episode 1|3 
TvShow 1 |Episode 2|3"

data <- read.csv(text = textFile,sep="|",stringsAsFactors = FALSE)
splitVar <- as.factor(data$Parent)
groupedData <- split(data,splitVar)

# second frame in list contains networks & shows
shows <- groupedData[[2]][-3]
colnames(shows) <- c("Network","Show")

# third frame in list contains shows and episodes
episodes <- groupedData[[3]][-3]
colnames(episodes) <- c("Show","Episode")

# use sqldf to join shows with episodes, since the shows data frame
# also includes the network names
library(sqldf)
sqlstmt <- "select s.Network, e.Show, e.Episode from shows s, episodes e where s.Show = e.Show"
result <- sqldf(sqlstmt)
result

...以及输出：

> result
    Network      Show   Episode
1 Network 1 TvShow 1  Episode 1
2 Network 1 TvShow 1  Episode 2
3 Network 1 TvShow 1  Episode 2
4 Network 1 TvShow 2  Episode 1
5 Network 1 TvShow 2  Episode 2
6 Network 1 TvShow 3  Episode 1
> 

dplyr版本

我们可以使用dplyr::inner_join()完成与sqldf()进行的数据帧连接。一旦将传入数据按Parent的值拆分为单独的数据帧，并从列表中提取出来以创建shows和episodes数据帧并重命名了列，我们就将两者合并数据帧如下。

# dplyr version
library(dplyr)
shows %>% inner_join(episodes, by = "Show")

...以及输出：

> shows %>% inner_join(episodes, by = "Show")
    Network      Show   Episode
1 Network 1 TvShow 1  Episode 1
2 Network 1 TvShow 1  Episode 2
3 Network 1 TvShow 1  Episode 2
4 Network 1 TvShow 2  Episode 1
5 Network 1 TvShow 2  Episode 2
6 Network 1 TvShow 3  Episode 1
> 

Answer 2

我猜这里最好的建议是将Docs分成2个不同的集合TVShows和Episodes，例如TvShows = filter(Docs, stringr::str_detect("TvShow"))

删除父列，调整列名称，然后full_join。