I wish to use pivot longer in the tibble below.

I wish to create tibble with 3 columns

1. Column 1 – name – alpha – contains a and b
2. Column 2 – name – beta – contains X and Y
3. Column 3 – name – P – contains values

library(tidyverse)

tbl <- tibble(b_X_P = runif(10),
              b_Y_P = runif(10),
              a_X_P = runif(10),
              a_Y_P = runif(10))

>Solution :

Here, we can just use names_sep as the _ will be delimiter where the column names should be split. Also, specify the names_to as a vector of 3 elements i.e. alpha, beta (as column names for the prefix part from the column names, and .value for the value part from the column)

library(tidyr)
pivot_longer(tbl, cols = everything(),
    names_to = c("alpha", "beta", ".value"), names_sep = "_")

-output

# A tibble: 40 × 3
   alpha beta        P
   <chr> <chr>   <dbl>
 1 b     X     0.271  
 2 b     Y     0.461  
 3 a     X     0.546  
 4 a     Y     0.344  
 5 b     X     0.234  
 6 b     Y     0.00462
 7 a     X     0.0157 
 8 a     Y     0.384  
 9 b     X     0.309  
10 b     Y     0.628  
# … with 30 more rows

---

If we need names_pattern it would be a pattern that should be wrapped within () to capture those characters

pivot_longer(tbl, cols = everything(),
    names_to = c("alpha", "beta", ".value"), 
    names_pattern = "^([^_]+)_([^_]+)_(.*)")
# A tibble: 40 × 3
   alpha beta        P
   <chr> <chr>   <dbl>
 1 b     X     0.271  
 2 b     Y     0.461  
 3 a     X     0.546  
 4 a     Y     0.344  
 5 b     X     0.234  
 6 b     Y     0.00462
 7 a     X     0.0157 
 8 a     Y     0.384  
 9 b     X     0.309  
10 b     Y     0.628  
# … with 30 more rows