R - 通过合并和超过 2 个后缀减少（或：如何合并多个数据帧并跟踪列）

Question

我正在尝试基于 2 列合并 4 个数据框，但要跟踪列源自哪个数据框。我在跟踪列时遇到了问题。

（见 dput(dfs) 文章结尾）

#df example (df1)
Name    Color    Freq
banana  yellow   3
apple   red      1
apple   green    4
plum    purple   8


#create list of dataframes
list.df <- list(df1, df2, df3, df4)

#merge dfs on column "Name" and "Color"
combo.df <- Reduce(function(x,y) merge(x,y, by = c("Name", "Color"), all = TRUE, accumulate=FALSE, suffixes = c(".df1", ".df2", ".df3", ".df4")), list.df)

这会给出以下警告：

警告信息： 在 merge.data.frame(x, y, by = c("Name", "Color"), all = TRUE, : 列名“Freq.df1”、“Freq.df2”在结果中重复

并输出此数据框：

#combo df example
Name    Color    Freq.df1   Freq.df2  Freq.df1  Freq.df2
banana  yellow   3          3         7         NA
apple   red      1          2         9         1
apple   green    4          NA        8         2
plum    purple   8          1         NA        6

df1 和 df2 仅在名称中重复。填充combo 的第三和第四列的值实际上分别来自df3 和df4。

我真正想要的是：

Name    Color    Freq.df1   Freq.df2  Freq.df3  Freq.df4
banana  yellow   3          3         7         NA
apple   red      1          2         9         1
apple   green    4          NA        8         2
plum    purple   8          1         NA        6

我怎样才能做到这一点？我知道 merge(..., suffixes) 函数只能处理 2 的字符向量，但我不知道应该如何解决。谢谢！

df1 <- 
structure(list(Name = structure(c(2L, 1L, 1L, 3L), .Label = c("apple", 
"banana", "plum"), class = "factor"), Color = structure(c(4L, 
3L, 1L, 2L), .Label = c("green", "purple", "red", "yellow"), class = "factor"), 
    Freq = c(3, 1, 4, 8)), .Names = c("Name", "Color", "Freq"
), row.names = c(NA, -4L), class = "data.frame")

df2 <-
structure(list(Name = structure(c(2L, 1L, 3L), .Label = c("apple", 
"banana", "plum"), class = "factor"), Color = structure(c(3L, 
2L, 1L), .Label = c("purple", "red", "yellow"), class = "factor"), 
    Freq = c(3, 2, 1)), .Names = c("Name", "Color", "Freq"), row.names = c(NA, 
-3L), class = "data.frame")

df3 <-
structure(list(Name = structure(c(2L, 1L, 1L), .Label = c("apple", 
"banana"), class = "factor"), Color = structure(c(3L, 2L, 1L), .Label = c("green", 
"red", "yellow"), class = "factor"), Freq = c(7, 9, 8)), .Names = c("Name", 
"Color", "Freq"), row.names = c(NA, -3L), class = "data.frame")

df4 <-
structure(list(Name = structure(c(1L, 1L, 2L), .Label = c("apple", 
"plum"), class = "factor"), Color = structure(c(3L, 1L, 2L), .Label = c("green", 
"purple", "red"), class = "factor"), Freq = c(1, 2, 6)), .Names = c("Name", 
"Color", "Freq"), row.names = c(NA, -3L), class = "data.frame")

Answer 1

使用for 循环似乎更容易，因为Reduce 或reduce (purrr) 一次只需要两个数据集，所以我们不能有超过两个suffixes 在merge.

在这里，我们创建了一个后缀向量 ('sfx')。使用第一个 list 元素初始化输出数据集。然后循环遍历“list.df”的序列，并在每个步骤中更新“res”时，使用“res”和list.df的下一个元素执行连续的merge

sfx <- c(".df1", ".df2", ".df3", ".df4")
res <- list.df[[1]]
for(i in head(seq_along(list.df), -1)) {

 res <- merge(res, list.df[[i+1]], all = TRUE, 
                 suffixes = sfx[i:(i+1)], by = c("Name", "Color"))
  }

res
#    Name  Color Freq.df1 Freq.df2 Freq.df3 Freq.df4
#1  apple  green        4       NA        8        2
#2  apple    red        1        2        9        1
#3 banana yellow        3        3        7       NA
#4   plum purple        8        1       NA        6

Answer 2

我终于可以使用Reduce 函数本身来完成这项工作。为此，我以特定格式修改了输入。

由于我们无法将data.frame 的名称作为参数传递给Reduce 函数，因此我创建了一个列表，其属性n 包含data.frame 的名称。

lst=list(list(n="df1",df=df1),list(n="df2",df=df2),list(n="df3",df=df3), list(n="df4",df=df4))

围绕这一点，我构建了一个逻辑来跟踪正在处理的data.frames 的名称。

Reduce(function(x,y){
    if(ncol(x$df)==3){
      #df column names after 1st merge.
      namecol=c('Name','Color',paste0("Freq.",x$n),paste0("Freq.",y$n))
    }else{
        #df column names for remaining merges.
        namecol=c(colnames(x$df),paste0("Freq.",y$n))
    }
    df=merge.data.frame(x = x$df,y = y$df,by = c("Name","Color"),all = TRUE)
    colnames(df)=namecol
    list(n="df",df=df)},lst)


#$n
#[1] "df"

#$df
#    Name  Color Freq.df1 Freq.df2 Freq.df3 Freq.df4
#1  apple  green        4       NA        8        2
#2  apple    red        1        2        9        1
#3 banana yellow        3        3        7       NA
#4   plum purple        8        1       NA        6

Answer 3

我的包safejoin的函数eat有这样的功能，如果你给 它是一个命名的 data.frames 列表作为第二个输入，它将加入它们 递归到带有此名称的新列前缀的第一个输入。 我们将不得不单独重命名。

# devtools::install_github("moodymudskipper/safejoin")
library(safejoin)
library(dplyr)
eat(rename(df1,df1_Freq = Freq), lst(df2,df3,df4),
    .by = c("Name","Color"), .mode= "full",.check="")
#     Name  Color df1_Freq df2_Freq df3_Freq df4_Freq
# 1 banana yellow        3        3        7       NA
# 2  apple    red        1        2        9        1
# 3  apple  green        4       NA        8        2
# 4   plum purple        8        1       NA        6

.mode = "full" 是进行完全外连接，虽然这里是默认的（左连接），但结果相同。

.check = "" 是删除检查，这会警告因素在连接列之间具有不同的级别。