dplyr join warning: joining factors with different levels

Question

When using the join function in the dplyr package, I get this warning:

Warning message:
In left_join_impl(x, y, by$x, by$y) :
  joining factors with different levels, coercing to character vector

There is not a lot of information online about this. Any idea what it could be? Thanks!

With errors like these, it helps to post a [reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) — MrFlick, May 26 '15 at 20:40

score 46 · Accepted Answer · edited Aug 14 '17 at 17:29

46

That's not an error, that's a warning. And it's telling you that one of the columns you used in your join was a factor and that factor had different levels in the different datasets. In order not to lose any information, the factors were converted to character values. For example:

library(dplyr)
x<-data.frame(a=letters[1:7])
y<-data.frame(a=letters[4:10])

class(x$a) 
# [1] "factor"

# NOTE these are different
levels(x$a)
# [1] "a" "b" "c" "d" "e" "f" "g"
levels(y$a)
# [1] "d" "e" "f" "g" "h" "i" "j"

m <- left_join(x,y)
# Joining by: "a"
# Warning message:
# joining factors with different levels, coercing to character vector 

class(m$a)
# [1] "character"

You can make sure that both factors have the same levels before merging

combined <- sort(union(levels(x$a), levels(y$a)))
n <- left_join(mutate(x, a=factor(a, levels=combined)),
    mutate(y, a=factor(a, levels=combined)))
# Joining by: "a"
class(n$a)
#[1] "factor"

edited Aug 14 '17 at 17:29

dfrankow

20,191
41
152
214

answered May 26 '15 at 20:40

MrFlick

195,160
17
277
295

1

sounds good, does that mean the code is good to go? Or are there some tweaks i need to make to get rid of the warning? – Christopher Yee May 26 '15 at 20:40
Well, you can make sure the factors have the same levels in each data set prior to the join. – MrFlick May 26 '15 at 20:41
1

they shouldnt have the same number of factors. I am basically joining a procedure table of about 20,000 codes to to a data set that may not have all those procedure codes present. – Christopher Yee May 26 '15 at 20:43
@EgorUfimtsev, instead of editing you should perhaps ask the answerer for clarification by a comment (concerning whether `left_join(mutate(x,` is correct, or it should read `left_join(mutate(y,` instead)... – aschipfl Aug 23 '16 at 18:30
1

Just reconvert to a factor once you have carried out the join. As @MrFlick said, it is a warning not an error. – Seanosapien Aug 24 '17 at 15:47

score 5 · Answer 2 · edited Jan 25 '19 at 11:11

5

This warning message will also appear if the joining columns in the two tables have different level orders;

tb1 <- data_frame(a = c("a","b","c")) %>% mutate(a=as.factor(a))
# Change level order of table tb2's col a
tb2 <- tb1 %>% mutate(a = fct_relevel(a,"c"))

# Check both still factors
tb1$a %>% class()
[1] "factor"
tb2$a %>% class()
[1] "factor"

# Check level order
tb1$a %>% levels()
[1] "a" "b" "c"
tb2$a %>% levels()
[1] "c" "a" "b"

# Try joining
tb1 %>% left_join(tb2)
Joining, by = "a"
Column `a` joining factors with different levels, coercing to character vector

edited Jan 25 '19 at 11:11

EcologyTom

2,344
2
27
38

answered Feb 28 '18 at 03:52

Jiaxiang

865
12
23

This was helpful. It seems the warning message is irrelevant in this case, the level orders shouldn't make a difference to the joining. – zola25 Dec 14 '18 at 14:28
@zola25 yes, the order is irrelavant, and this is one of examples with the kind of warning. – Jiaxiang Dec 18 '18 at 05:25

score 2 · Answer 3 · answered Nov 05 '18 at 16:29

2

In case of database, don't forget stringsAsFactors=FALSE in many cases, to avoid this warning. (It's was my case).

sqlExecute(my_database_channel, data=myparam, stringsAsFactors=FALSE )

answered Nov 05 '18 at 16:29

phili_b

885
9
27

dplyr join warning: joining factors with different levels

3 Answers3

Linked