CODEWITHSUNDEEP
rtype-conversiontidyversedplyr
rainbird
rainbird

Reputation: 337

Converting double to integers using R tidyverse

Following How to convert column types in R tidyverse I am trying to convert doubles (numeric) into integers.

For example, using the iris data:

iris1 <- iris %>%
  mutate_at(vars(Petal.Length), integer)

The above throws an error, which I cannot understand despite following the recommended trouble shooting:

Error: Problem with `mutate()` column `Petal.Length`.
ℹ `Petal.Length = (function (length = 0L) ...`
.x invalid 'length' argument

Using the same line of code to convert to factor and the results are fine:

iris1 <- iris %>%
mutate_at(vars(Petal.Length), factor)
class(iris1$Petal.Length)

Can someone explain to me the reason for the error and how I can convert to integer. Ideally I am looking for a solution friendly to the pipe operator %>%.

Upvotes: 2

Views: 7698

Answers (2)

Gregory McFann
Gregory McFann

Reputation: 1

I ran into the same problem trying to run "uncount" on a column that had been converted from a percentage to an absolute number, so it was in effect an integer but still technically a double

# A tibble: 18 × 2
   MobileOS      OS_2010
   <fct>           <dbl>
 1 Android          1190
 2 BlackBerry OS    2013
 3 Brew                0
 4 iOS              5358
 5 LG                 52
 6 Linux               0
 7 Nintendo           79
 8 Nintendo 3DS        0
 9 Nokia Unknown       0
10 Other              77
11 Playstation       218
12 Samsung           174
13 Series 40           0
14 Sony Ericsson      43
15 SymbianOS         141
16 Unknown           433
17 webOS             117
18 Windows           105

Apparently you could get away with that in the old days

So I tried the older way to coerce the double to integer

df %<>%
  mutate_at(vars(OS_2010), as.integer) %>%
  print()

That worked, and the "uncount" worked

Then I reran the problem and used the modern way

df %<>%
  mutate(across(
    .cols = matches('OS_2010'),
    .fns = ~ as.integer(.x)))

That worked also.

Upvotes: 0

Ian Cero
Ian Cero

Reputation: 370

I think the problem you are running into is that integer() creates an integer-typed vector, which is just an empty place-holder for integers to go into. What you want is as.integer(), which takes an existing numeric vector and coerces each element into an integer. It's a minor distinction, so it's easy to miss. Here is your code with the fix implemented:

A technically correct answer

iris1 <- iris %>%
  mutate_at(vars(Petal.Length), as.integer) # note the AS.INTERGER()

A more modern approach

You might be interested to know that mutate_at() has been superseded by the more modern mutate(across(...))) approach. Depending on your goals, you might prefer the more modern way to solve this problem with dplyr:

iris2 <- iris %>% 
  mutate(across(
    .cols = matches('Petal.Length'),
    .fns = ~ as.integer(.x)))

Don't forget about rounding

Just a heads up, coercing something into an integer drops any decimal values, which is basically like rounding down every value to the nearest (lower) integer. You might want to round your numbers first, then coerce them to integer format, depending on your other goals for this code.

iris1 <- iris %>%
  mutate_at(vars(Petal.Length), ~ as.integer(round(.x)))

OR

iris2 <- iris %>% 
  mutate(across(
    .cols = matches('Petal.Length'),
    .fns = ~ round(as.integer(.x))))

Upvotes: 3

Related Questions