For when someone asked you to do something you've done before, you can argue that the quickest way to do it is to just take the work someone else did and utilize that. No reason to reinvent the wheel.
Details
Multiples of 3 are shown as "Fizz"; multiples of 5 as
"Buzz"; multiple of both (i.e., 15) are "FizzBuzz".
fizzbuzz_lazy() subsets the .fizzbuzz_vector object, which is a
solution with default parameters up to 1e6
Examples
fizzbuzz(15)
#> [1] "1" "2" "Fizz" "4" "Buzz" "Fizz"
#> [7] "7" "8" "Fizz" "Buzz" "11" "Fizz"
#> [13] "13" "14" "FizzBuzz"
fizzbuzz(30, show_numbers = FALSE)
#> [1] "" "" "Fizz" "" "Buzz" "Fizz"
#> [7] "" "" "Fizz" "Buzz" "" "Fizz"
#> [13] "" "" "FizzBuzz" "" "" "Fizz"
#> [19] "" "Buzz" "Fizz" "" "" "Fizz"
#> [25] "Buzz" "" "Fizz" "" "" "FizzBuzz"
cat(fizzbuzz(30), sep = "\n")
#> 1
#> 2
#> Fizz
#> 4
#> Buzz
#> Fizz
#> 7
#> 8
#> Fizz
#> Buzz
#> 11
#> Fizz
#> 13
#> 14
#> FizzBuzz
#> 16
#> 17
#> Fizz
#> 19
#> Buzz
#> Fizz
#> 22
#> 23
#> Fizz
#> Buzz
#> 26
#> Fizz
#> 28
#> 29
#> FizzBuzz
# \donttest{
# show them how fast your solution is:
if (fuj::available_namespace("bench")) {
bench::mark(fizzbuzz(1e5), fizzbuzz_lazy(1e5))
}
#> # A tibble: 2 × 13
#> expression min median `itr/sec` mem_alloc `gc/sec` n_itr n_gc total_time
#> <bch:expr> <bch:t> <bch:t> <dbl> <bch:byt> <dbl> <int> <dbl> <bch:tm>
#> 1 fizzbuzz(… 40.6ms 41.1ms 24.1 5.19MB 4.81 10 2 416ms
#> 2 fizzbuzz_… 615.1µs 634.8µs 1542. 38.21MB 53.9 372 13 241ms
#> # ℹ 4 more variables: result <list>, memory <list>, time <list>, gc <list>
# }