13

I am learning Elixir and found myself having to concat multiple lists together where I have a beginning, middle, and end. Simplified example:

a = [1,2]
b = [3,4]
c = [5,6]
a ++ b ++ c
> [1, 2, 3, 4, 5, 6]

I am doing this thousands of times in a stream and want to be Elixirific about it.

Part 1)

I wrote a function to handle this, there is probably something that does this for me, but I don't see it. Should I be using built-in Elixir function for this?

def append(front, back) when is_list(front) when is_list(back) do
  front
  |> Enum.reverse
  |> Enum.reduce(back, &([&1 | &2]))
end

Or should I have just done this and it will become more natural for me over time?

[1, 2]
|> Enum.reverse 
|> Enum.reduce([3, 4], &([&1 | &2]))
|> Enum.reverse
|> Enum.reduce([5, 6], &([&1 | &2]))

Part 2)

Does the order in which I call the parts together matter?

Way 1:
[1, 2]
|> append([3, 4])
|> append([5, 6])

...
Way 2:
end = append([3, 4], [5, 6])
append([1, 2], end)

Will the intermediary list be re-used in both scenarios since both are appending header pointers?

Any help on this would be great, cheers.

1 Answer 1

23

No matter what approach you use, you will end up at cloning at least all lists except the last one, since you cannot append to a linked list without cloning it. So if you know you want to concatenate the lists (i.e. you cannot modify your code to accept a nested list like [a, b, c]), I would suggest a ++ b ++ c since ++ is implemented in Erlang's C code, which should be pretty much as fast as possible, and definitely faster than manually concatenating with Enum.reverse/1 and Enum.reduce/3.

Here's a tiny microbenchmark comparing a ++ b ++ c vs a |> append(b) |> append(c):

defmodule BasicBench do
  use Benchfella

  bench "++", [a: gen(), b: gen(), c: gen()] do
    a ++ b ++ c
  end

  bench "append", [a: gen(), b: gen(), c: gen()] do
    a |> append(b) |> append(c)
  end

  def append(front, back) when is_list(front) when is_list(back) do
    front
    |> Enum.reverse
    |> Enum.reduce(back, &([&1 | &2]))
  end

  def gen, do: Enum.to_list(1..1000)
end

and the output:

benchma iterations   average time
++          100000   10.22 µs/op
append       20000   75.79 µs/op
Sign up to request clarification or add additional context in comments.

1 Comment

Sidenote: there is also Enum.concat/1 that performs slightly better than manual reduce, but still ≈5 times worse than plain old good :erlang.++. On the other hand, Enum.concat works for any enumerable.

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.