如何使用正则表达式将 data.frame 重塑为具有多个值列的长格式

Question

How do I reshape the data.frame input to result ? Basically the first part of the column name before "dosis" should be the new variable with two value columns value and dosis containing the data of the columns ending with/without "dosis" .

This should not be too difficult, but I have difficulties finding the correct regex to use with pivot_longer or melt.data.table .

library(tibble)
library(tidyr)
library(magrittr)
library(data.table)

input <-
  tribble(
    ~"abc", ~"abcdosis", ~"def", ~"defdosis", ~"ghi", ~"ghidosis",
    1, 0, 9, NA, 1, 2
  )

result  <-
  tribble(
    ~"variable", ~"value", ~"dosis",
    "abc", 1, 0,
    "def", 9, NA,
    "ghi", 1, 2
  )

# Not working
pivot_longer(input, 
             everything(), 
             names_to = c("variable", "dosis"),
             names_pattern = "(^dosis)?(dosis)")

# Also not working
melt.data.table(as.data.table(input), measure.vars = patterns("^(?!.*dosis).*$", "dosis$"))

Answer 1

Using dplyr::rename_with() you can paste "value" onto the end of of the non- "dosis" columns, then use then use the ".value" sentinel in pivot_longer() . Of course, may have to be more specific with the columns in rename_with() if you data is more complex.

library(dplyr)
library(tidyr)

input %>%
  rename_with(~paste0(., "value"), -ends_with("dosis")) %>%
  pivot_longer(everything(), names_to = c("variable", ".value"), names_pattern = "(.*?)(value|dosis)$")

# A tibble: 3 x 3
  variable value dosis
  <chr>    <dbl> <dbl>
1 abc          1     0
2 def          9    NA
3 ghi          1     2