pivot_longer with names_pattern and pairs of columns

January 13, 2022

I am trying to figure out how to use pivot_longer from tidyr in the following example. This is how the original table called dat_plot is structured like:

   year organizational_based action_based ideological_based share_org_based share_ideo_based share_act_based
  <dbl>                <dbl>        <dbl>             <dbl>           <dbl>            <dbl>           <dbl>
1  1956                    1            0                 0               2               95              95
2  2000                    0            0                 0              92               87              91

also here:

dat_plot <- structure(list(year = c(1956, 2000), organizational_based = c(1, 
0), action_based = c(0, 0), ideological_based = c(0, 0), share_org_based = c(2, 
92), share_ideo_based = c(95, 87), share_act_based = c(95, 91
)), row.names = c(NA, -2L), class = c("tbl_df", "tbl", "data.frame"
))

I would like to turn it into the long format, in the following way:

  year          based based_value      share share_value
1 1956 organizational           1  org_based           2
2 1956         action           0 ideo_based          95
3 1956    ideological           0  act_based          95
4 2000 organizational           0  org_based          92
5 2000         action           0 ideo_based          87
6 2000    ideological           0  act_based          91

or, with dput:

solution <- structure(list(year = c(1956, 1956, 1956, 2000, 2000, 2000), 
    based = c("organizational", "action", "ideological", "organizational", 
    "action", "ideological"), based_value = c(1, 0, 0, 0, 0, 
    0), share = c("org_based", "ideo_based", "act_based", "org_based", 
    "ideo_based", "act_based"), share_value = c(2, 95, 95, 92, 
    87, 91)), class = "data.frame", row.names = c(NA, -6L))

I thought I have to work with names_pattern, what I have tried is something like this, but if you try you will see, that it is not what I want:

pivot_longer(data=dat_plot, cols=c("share_org_based", "share_ideo_based", "share_act_based",
                    "organizational_based", "action_based", "ideological_based"),
             names_pattern = c("(share_[A-Za-z]+)([A-Za-z]+_based)"),
             names_to = c("share", ".value"),
             values_to = "value")

I am grateful for any leads as to how names_pattern works, or what am I missing.

>Solution :

Here’s another way with tidyr:

dat_plot %>% 
  pivot_longer(cols = starts_with("share"), names_to = "share", names_prefix = "share_", values_to = "share_value") %>%
  pivot_longer(cols = ends_with("based"), names_to = "based", names_pattern = "(.*)_based", values_to = "based_value") %>% 
  filter(substr(share, 1, 3) == substr(based, 1, 3))

output

# A tibble: 6 x 5
   year share      share_value based          based_value
  <dbl> <chr>            <dbl> <chr>                <dbl>
1  1956 org_based            2 organizational           1
2  1956 ideo_based          95 ideological              0
3  1956 act_based           95 action                   0
4  2000 org_based           92 organizational           0
5  2000 ideo_based          87 ideological              0
6  2000 act_based           91 action                   0