86

I'm trying to mutate a new variable from sort of row calculation, say rowSums as below

iris %>% 
  mutate_(sumVar = 
            iris %>% 
            select(Sepal.Length:Petal.Width) %>%
            rowSums)

the result is that "sumVar" is truncated to its first value(10.2): 

Source: local data frame [150 x 6]
Groups: <by row>

   Sepal.Length Sepal.Width Petal.Length Petal.Width Species sumVar
1           5.1         3.5          1.4         0.2  setosa   10.2
2           4.9         3.0          1.4         0.2  setosa   10.2
3           4.7         3.2          1.3         0.2  setosa   10.2
4           4.6         3.1          1.5         0.2  setosa   10.2
5           5.0         3.6          1.4         0.2  setosa   10.2
6           5.4         3.9          1.7         0.4  setosa   10.2
..
Warning message:
Truncating vector to length 1

Should it be rowwise applied? Or what's the right verb to use in these kind of calculations.

Edit:

More specifically, is there any way to realize the inline custom function with dplyr?

I'm wondering if it is possible do something like:

iris %>% 
  mutate(sumVar = colsum_function(Sepal.Length:Petal.Width))
Improve this question
3
  • 6
    Really strange that iris %>% select(Sepal.Length:Petal.Width) %>% rowSums() works fine but iris %>% mutate(sumVar = iris %>% select(Sepal.Length:Petal.Width) %>% rowSums()) throws a "Error: Bad indices 1" + warning message. Commented Dec 8, 2014 at 9:38
  • I am trying to work on it with different approaches, but this error appears very frequently using . (I am also doing something silly sometimes). Commented Dec 8, 2014 at 9:49
  • 4
    For operations like sum that already have an efficient vectorised row-wise alternative, the proper way is currently: df %>% mutate(total = rowSums(across(where(is.numeric)))) across can take anything that select can (e.g. rowSums(across(Sepal.Length:Petal.Width)) also works). See the full spiel about row-wise and across Commented Apr 15, 2021 at 8:46

7 Answers 7

Reset to default
146

This is more of a workaround but could be used

iris %>% mutate(sumVar = rowSums(.[1:4]))

As written in comments, you can also use a select inside of mutate to get the columns you want to sum up, for example

iris %>% 
  mutate(sumVar = rowSums(select(., contains("Sepal")))) %>% 
  head

or 

iris %>% 
  mutate(sumVar = select(., contains("Sepal")) %>% rowSums()) %>% 
  head
Improve this answer
Sign up to request clarification or add additional context in comments.

7 Comments

Which version of dplyr are you using? When I try you example with dplyr_0.4.1, I receive an exception: Error in is.data.frame(x) : object '.' not found.
If it's of use to anyone, the reason why I was receiving the error Error in is.data.frame(x) : object '.' not found was because I had an old version of magrittr. When I updated from magrittr_1.0.1 to magrittr_1.5, everything worked fine.
@Konrad, you could do something like iris %>% mutate(sumVar = rowSums(select(., contains("Sepal")))) %>% head or iris %>% mutate(sumVar = select(., contains("Sepal")) %>% rowSums()) %>% head
The comment by @docendodiscimus really should be another (vote-able) answer. It is the most robust dplyr-esque solution.
It's nice that this works, although Hadley says that a solution like this "works by coincidence, not by design. I wouldn't rely on it." But maybe it is supported now? Does anyone know? github.com/tidyverse/dplyr/issues/2050
|
26

You can use rowwise() function:

iris %>% 
  rowwise() %>% 
  mutate(sumVar = sum(c_across(Sepal.Length:Petal.Width)))

#> # A tibble: 150 x 6
#> # Rowwise: 
#>    Sepal.Length Sepal.Width Petal.Length Petal.Width Species sumVar
#>           <dbl>       <dbl>        <dbl>       <dbl> <fct>    <dbl>
#>  1          5.1         3.5          1.4         0.2 setosa    10.2
#>  2          4.9         3            1.4         0.2 setosa     9.5
#>  3          4.7         3.2          1.3         0.2 setosa     9.4
#>  4          4.6         3.1          1.5         0.2 setosa     9.4
#>  5          5           3.6          1.4         0.2 setosa    10.2
#>  6          5.4         3.9          1.7         0.4 setosa    11.4
#>  7          4.6         3.4          1.4         0.3 setosa     9.7
#>  8          5           3.4          1.5         0.2 setosa    10.1
#>  9          4.4         2.9          1.4         0.2 setosa     8.9
#> 10          4.9         3.1          1.5         0.1 setosa     9.6
#> # ... with 140 more rows

"c_across() uses tidy selection syntax so you can to succinctly select many variables"'

Finally, if you want, you can use %>% ungroup at the end to exit from rowwise.

Improve this answer

3 Comments

For operations like sum that already have an efficient vectorised row-wise alternative, the proper way is currently: df %>% mutate(total = rowSums(across(where(is.numeric)))) across can take anything that select can (e.g. rowSums(across(Sepal.Length:Petal.Width)) also works). Scroll down the row-wise vignette to find this and have a look at across
@FonsMA this should rather be an answer, and even an accepted answer, than a comment
@DanChaltiel, thanks, I didn't see the comment had this many upvotes! it's now an answer below
17

As requested, transforming my commment into an answer:

For operations like sum that already have an efficient vectorised row-wise alternative, the proper way is currently:

df %>% mutate(total = rowSums(across(where(is.numeric))))

across can take anything that select can (e.g. rowSums(across(Sepal.Length:Petal.Width)) also works).

Scroll down the row-wise vignette to find this and have a look at across

Improve this answer

Comments

12

A more complicated way would be:

 iris %>% select(Sepal.Length:Petal.Width) %>%
mutate(sumVar = rowSums(.)) %>% left_join(iris)
Improve this answer

2 Comments

Thanks Davide. left_join sounds a nice solution if using it with by key; however, it's not so robust and intuitive for this circumstance
I also worry the automatic "by" parameter selection in the join could cause some troubles.. the columns could contain non-unique values on some rows ..
11

Adding @docendodiscimus's comment as an answer. +1 to him!

iris %>% mutate(sumVar = rowSums(select(., contains("Sepal"))))
Improve this answer

Comments

3

I am using this simple solution, which is a more robust modification of the answer by Davide Passaretti:

iris %>% select(Sepal.Length:Petal.Width) %>%
  transmute(sumVar = rowSums(.)) %>% bind_cols(iris, .)

(But it requires a defined row order, which should be fine, unless you work with remote datasets perhaps..)

Improve this answer

2 Comments

Hi could you please tell me the meaning of dot in between round brackets? rowSums(.)
The dot represents the result coming out of the %>% pipe
1

You can also use a grep in place of contains or matches, just in case you need to get fancy with the regular expressions (matches doesn't seem to much like negative lookaheads and the like in my experience).

iris %>% mutate(sumVar = rowSums(select(., grep("Sepal", names(.)))))
Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.