Count Distinct values from 2 columns - r

Question

I need to get only distinct values that are spread over two columns and return the distinct values into one column.
Example:
colA colB
---- --------
darcy elizabeth
elizabeth darcy
jon doe
doe joe
It should return:
resultCol
darcy
elizabeth
jon
doe
Is there any builtin function or library that can do that more efficiently?
I tried a workaround to get the results but it is extremely slow for more than 100 thousands observations.
#First i create a sample dataframe
col1<-c("darcy","elizabeth","elizabeth","darcy","john","doe")
col2<-c("elizabeth","darcy","darcy","elizabeth","doe","john")
dfSample<-data.frame(col1,col2)
#Then i create an empty dataframe to store all values in a single column
emptyDataframe<-data.frame(resultColumn=character())
for(i in 1:nrow(dfSample)){
emptyDataframe<-rbind(emptyDataframe,c(toString(dfSample[i,1])),stringsAsFactors=FALSE)
}
for(i in 1:nrow(dfSample)){
emptyDataframe<-rbind(emptyDataframe,c(toString(dfSample[i,2])),stringsAsFactors=FALSE)
}
emptyDataframe
#Finally i get the distinct values using dplyr
var_distinct_values<-distinct(emptyDataframe)

Answer 1

I use union to get unique values across specific columns:
with(dfSample, union(col1, col2))
PS: The answer from d.b in the comments is also another way.
You can improvise his answer if you have extra columns but want to run it only over specific columns:
unique(unlist(dfSample[1:2]))
This gets the unique values from first two columns.

Answer 2

Here is a general purpose solution.
It's based on this answer but can be extended to any number of columns as long as the object is a data.frame or list.
Reduce(union, dfSample)
[1] "darcy" "elizabeth" "john" "doe"
Now with 100K observations in each of 10 columns.
set.seed(1234)
n <- 1e5
bigger <- replicate(n, sample(c(col1, col2), 10, TRUE))
bigger <- as.data.frame(bigger)
system.time(Reduce(union, bigger))
# user system ellapsed
# 3.769 0.000 3.772
Edit.
After a second thought, I realized that the test above is run with a dataframe with a very small number of different values. A test with a larger number does not necessarily give the same results.
set.seed(1234)
s <- sprintf("%05d", 1:5000)
big2 <- replicate(n, sample(s, 10, TRUE))
big2 <- as.data.frame(big2)
rm(s)
microbenchmark::microbenchmark(
red = Reduce(union, big2),
uniq = unique(unlist(big2)),
times = 10
)
#Unit: seconds
# expr min lq mean median uq max neval cld
# red 26.021855 26.42693 27.470746 27.198807 28.56720 29.022047 10 b
# uniq 1.405091 1.42978 1.632265 1.548753 1.56691 2.693431 10 a
The unique/unlist solution is now clearly better.