如何融化具有多个组和空后缀的data.frame

Question

我有一个data.frame，需要根据列的后缀将多个列融合在一起。因此，所有以“from”结尾的列都应该合并为一列，对于以“to”结尾的列和所有没有后缀的列都一样，这就是我的问题，因为如果不附加没有后缀的列名，我就无法融合“xxx”。问题是如何将空后缀与names_pattern 中的正则表达式匹配，或者是否有不重命名的不同解决方案？我也对data.table这个问题的解决方案感兴趣。

library(tibble)
library(magrittr)
library(tidyr)

data <-
  tibble::tribble(
  ~"abc", ~"abcfrom", ~"abcto", ~"def", ~"deffrom", ~"defto",
  1, "2019-05-16", NA, 0, NA, NA,
  1, "2020-01-01", "2020-10-15", 1, "2014-12-17", "2015-03-05",
  1, NA, NA, 1, "2015-01-01", NA
)

data %>% 
  dplyr::rename("abcxxx" = "abc", "defxxx" = "def") %>% 
  tidyr::pivot_longer(
    everything(),
    names_to = c("variable", ".value"),
    names_pattern = "(.+)(xxx|from|to)"
  )

# A tibble: 6 x 4
  variable   xxx from       to        
  <chr>    <dbl> <chr>      <chr>     
1 abc          1 2019-05-16 NA        
2 def          0 NA         NA        
3 abc          1 2020-01-01 2020-10-15
4 def          1 2014-12-17 2015-03-05
5 abc          1 NA         NA        
6 def          1 2015-01-01 NA

Answer 1

这是一个选项

library(dplyr)
library(tidyr)
library(stringr)
data %>% 
 rename_at(vars(names(.)[!str_detect(names(.), "(from|to)$")]),
        ~ str_c(., 'xxx')) %>% 
 tidyr::pivot_longer(
 everything(),
 names_to = c("variable", ".value"),
 names_pattern = "(.+)(xxx|from|to)"
)

-输出

# A tibble: 6 x 4
#  variable   xxx from       to        
#  <chr>    <dbl> <chr>      <chr>     
#1 abc          1 2019-05-16 <NA>      
#2 def          0 <NA>       <NA>      
#3 abc          1 2020-01-01 2020-10-15
#4 def          1 2014-12-17 2015-03-05
#5 abc          1 <NA>       <NA>      
#6 def          1 2015-01-01 <NA>      

Answer 2

这是data.table 方法。

它并没有真正融化，而是根据列名的前三个相同字符分成（三）列的块。然后设置块的列名，再次将它们行绑定在一起..

library( data.table )
setDT( data )

#assuming the first three characters define the group
data.split <- split.default(data, gsub('(^.{3}).*$', '\\1', names(data)))
# $abc
#    abc    abcfrom      abcto
# 1:   1 2019-05-16       <NA>
# 2:   1 2020-01-01 2020-10-15
# 3:   1       <NA>       <NA>
#   
# $def
#    def    deffrom      defto
# 1:   0       <NA>       <NA>
# 2:   1 2014-12-17 2015-03-05
# 3:   1 2015-01-01       <NA>

#set column names (fixed here, but can also be variable if desired)
data.split <- lapply( data.split, setnames, new = c("xxx", "from", "to") )

#bind together
DT <- rbindlist( data.split, use.names = TRUE, idcol = "variable" )
#    variable xxx       from         to
# 1:      abc   1 2019-05-16       <NA>
# 2:      abc   1 2020-01-01 2020-10-15
# 3:      abc   1       <NA>       <NA>
# 4:      def   0       <NA>       <NA>
# 5:      def   1 2014-12-17 2015-03-05
# 6:      def   1 2015-01-01       <NA>