Gathering data by paired columns - r

Question

I'm having trouble in shaping my dataframe.
Here's an example:
id institution name1 id1 name2 id2
1 usp Miles Davis 123 Arturo Sandoval 111
2 unb Chet Baker 321 Clifford Brown 121
3 usp Wayne Shorter 222 Hermeto Pascoal 322
4 Puc-rio John Coltrane 333 Charlie Parker 112
I need to keep the id and institution columns and gather the other ones like this:
id institution name_all id_all
1 usp Miles Davis 123
1 usp Arturo Sandoval 111
2 unb Chet Baker 321
2 unb Clifford Brown 121
3 usp Wayne Shorter 222
3 usp Hermeto Pascoal 322
4 Puc-rio John Coltrane 333
4 Puc-rio Charlie Parker 112
I'm using the gather function from the dplyr:
df %>%
gather(name_all, id_all, -id, -institution)
but it comes like this:
id institution name id
1 usp name1 Miles Davis
1 usp id1 123
2 unb name1 Chet Baker
2 unb id2 121
Any ideas on how to pair those values? I have more than 5 columns to do so, I think that I'm missing an argument to specify which one of them are paired. I hope I've made myself clear.

Answer 1

For a tidyverse solution, you can:
library(dplyr)
library(tidyr)
df %>%
gather(ColType, ColValue, -id, -institution) %>%
mutate(id_number = gsub("^(\\D*)(\\d*)$", "\\2", ColType, ignore.case = TRUE, perl = TRUE),
ColType = gsub("^(\\D*)(\\d*)$", "\\1", ColType, ignore.case = TRUE, perl = TRUE)
) %>%
spread(ColType, ColValue) %>%
select(-id_number)

Answer 2

I'm sure that there is a more elegant solution, but you can try:
df %>%
gather(var, name_all, -matches("id|institution")) %>%
gather(var2, val, -c(id, institution, var, name_all)) %>%
mutate(id_all = ifelse(parse_number(var) == parse_number(var2), val, NA)) %>%
na.omit() %>%
select(-var, -var2, -val) %>%
arrange(id)
id institution name_all id_all
1 1 usp Miles_Davis 123
2 1 usp Arturo_Sandoval 111
3 2 unb Chet_Baker 321
4 2 unb Clifford_Brown 121
5 3 usp Wayne_Shorter 222
6 3 usp Hermeto_Pascoal 322
7 4 Puc-rio John_Coltrane 333
8 4 Puc-rio Charlie_Parker 112
First, it transforms the data from wide to long, excluding the variables that are named institution or id. Second, it performs a second wide-to-long transformation to have all the numbered "id" variables and their values as separate rows. Third, it checks whether the "name" variable has the number as the "id variable. If so, it assigns the appropriate value, otherwise NA. Finally, it removes the rows with NAs, the redundant variables and arranges the data.
Sample data:
df <- read.table(text = "
id institution name1 id1 name2 id2
1 usp Miles_Davis 123 Arturo_Sandoval 111
2 unb Chet_Baker 321 Clifford_Brown 121
3 usp Wayne_Shorter 222 Hermeto_Pascoal 322
4 Puc-rio John_Coltrane 333 Charlie_Parker 112", header = TRUE, stringsAsFactors = FALSE)