CODEWITHSUNDEEP
rtidyrtidyverse
Mike
Mike

Reputation: 2097

Combining Multiple Columns with Tidyr's Unite by Referencing Similar Column Names

library(tidyr)
library(dplyr)
library(tidyverse)

Below is the code for a simple dataframe. I have some messy data that was exported with column factor categories spread out in different columns. 

Client<-c("Client1","Client2","Client3","Client4","Client5")
Sex_M<-c("Male","NA","Male","NA","Male")
Sex_F<-c(" ","Female"," ","Female"," ")
Satisfaction_Satisfied<-c("Satisfied"," "," ","Satisfied","Satisfied")
Satisfaction_VerySatisfied<-c(" ","VerySatisfied","VerySatisfied"," "," ")
CommunicationType_Email<-c("Email"," "," ","Email","Email")
CommunicationType_Phone<-c(" ","Phone ","Phone "," "," ")
DF<-tibble(Client,Sex_M,Sex_F,Satisfaction_Satisfied,Satisfaction_VerySatisfied,CommunicationType_Email,CommunicationType_Phone)

I want to recombine the categories into single columns using tidyr's "unite". 

DF<-DF%>%unite(Sat,Satisfaction_Satisfied,Satisfaction_VerySatisfied,sep=" ")%>%
unite(Sex,Sex_M,Sex_F,sep=" ")

However, I have to write multiple "unite" lines and I feel this violates the three times rule, so there must be a way to make this easier, especially since my real data contains dozens of columns that need to be combined. Is there a way to use "unite" once but somehow refer to matching column names so that all column names that are similar (For example, containing "Sex" for "Sex_M" and "Sex_F", and "CommunicationType" for "CommunicationType_Email" and "CommunicationType_Phone") are combined with the above formula?

I was also thinking about a function that allows me to enter column names, but this is too difficult for me since it involves complex standard evaluation.

Upvotes: 3

Views: 2666

Answers (2)

akrun
akrun

Reputation: 887691

We can use unite

library(tidyverse)
DF %>% 
    unite(Sat, matches("^Sat"))

For multiple cases, perhaps

gather(DF, Var, Val, -Client, na.rm = TRUE) %>%
        separate(Var, into = c("Var1", "Var2")) %>%
        group_by(Client, Var1) %>% 
        summarise(Val = paste(Val[!(is.na(Val)|Val=="")], collapse="_")) %>%
        spread(Var1, Val)
#  Client CommunicationType  Satisfaction    Sex
#*   <chr>             <chr>         <chr>  <chr>
#1 Client1             Email     Satisfied   Male
#2 Client2             Phone VerySatisfied Female
#3 Client3             Phone VerySatisfied   Male
#4 Client4             Email     Satisfied Female
#5 Client5             Email     Satisfied   Male

Upvotes: 7

Jean
Jean

Reputation: 1490

Something like this? If you have loads of columns.

result<-with(new.env(),{
  Client<-c("Client1","Client2","Client3","Client4","Client5")
  Sex_M<-c("Male","NA","Male","NA","Male")
  Sex_F<-c(" ","Female"," ","Female"," ")
  Satisfaction_Satisfied<-c("Satisfied"," "," ","Satisfied","Satisfied")
  Satisfaction_VerySatisfied<-c(" ","VerySatisfied","VerySatisfied"," "," ")
  CommunicationType_Email<-c("Email"," "," ","Email","Email")
  CommunicationType_Phone<-c(" ","Phone ","Phone "," "," ")
  x<-ls()
  categories<-unique(sub("(.*)_(.*)", "\\1", x))
  df<-setNames(data.frame( lapply(x, function(y) get(y))), x)
  for(nm in categories){
    df<-unite_(df, nm, x[contains(vars = x, match = nm)])
  }
  return(df)
})

Client CommunicationType    Satisfaction       Sex
1 Client1           Email_      Satisfied_      _Male
2 Client2           _Phone   _VerySatisfied Female_NA
3 Client3           _Phone   _VerySatisfied     _Male
4 Client4           Email_      Satisfied_  Female_NA
5 Client5           Email_      Satisfied_      _Male

Upvotes: 0

Related Questions