libjohn / codestub-full_join-separate-for-loop

an example of how to transform some data using tidyverse tools

Simplified tidyverse approach with full_join (i.e. outer join)

side_a <- country_df %>% 
  filter(side_a == 0)

side_b <- country_df %>% 
  filter(side_a == 1)
  
full_join(side_b, side_a, by = "dispute") %>% 
  select(dispute, country.x, country.y)

Details:

Part of a project Mara brought to Ryan and me. Goal is to transform

Country	Dispute	Side A	Side B
A	1	0	1
B	1	0	1
C	1	1	0
D	1	1	0
A	2	0	1
E	2	1	0
F	2	1	0

to

Country.x	Country.y	Dispute
A	C	1
B	C	1
A	D	1
B	D	1
A	E	2
A	F	2

First Example took me down an long (and incorrect) road because my logic branched into the gutter. I found myself using a for loop with also requried the use of separate to transform multi-valed cells into rows.

While that was an interesting approach. Ryan's outer join approach was MUCH more elegant and consice. In tidyverse, outer join is full_join(). Once back on track, it became three simple setps.

The For Loop and the separate function

Well, I got tripped up on this originally. My first script goes round and round in a big old for-loop circle. But all was not for naught since I forced myself to learn some techniques which may come in handy later.

Below are the highlights using for and separate

# declare Tibble (data frame)
df_total = tibble()

for (i in 1:max(country_df$dispute)) {
  # Build vector of disputants (multi-valued cell: country_y) 
  country_y <- country_df %>% 
    filter(dispute == i) %>% 
    select(-dispute) %>% 
    mutate(side_b = if_else(side_b == 0, country, NA_character_)) %>% 
    select(side_b) %>% 
    drop_na() %>% 
    unique() %>% 
    as_vector() %>% 
    paste0(collapse = "|") 
  
  # Build a data frame that adds the disputants vector as a column-variable; 
  # drops blank rows ; separate the multi-valued cells (disputants/country_y)
  # into rows
  country_df_temp <- country_df %>% 
    filter(dispute == i) %>% 
    mutate(country_disputes = if_else(side_b == 1, country_y, NA_character_)) %>% 
    drop_na() %>% 
    separate_rows(country_disputes, convert = TRUE) 
  
  # Append big Dataframe from from the iterated dataframes
  df_total <- bind_rows(df_total, country_df_temp)
  
}

df_total %>%
  arrange(country_disputes) %>% 
  rename(country_x = country,
         counrty_y = country_disputes) %>% 
  select(dispute, counrty_y, country_x)