Elixir: Stream.drop_while generalisation? - stream

Question

I'd like to generalise this code but can't see how to do so in a neat way:
defmodule Demo do
defp side_effect(bool, label) do
if bool do
IO.puts label
end
end
def go do
{a,b,c} = {4,8,13}
[2,3,4,5,7,8,9,10,11,12] # always in ascending order
|> Stream.drop_while( fn i -> bool = (i<a); side_effect(bool, "a_#{i}"); bool end )
|> Stream.drop_while( fn i -> bool = (i<b); side_effect(bool, "b_#{i}"); bool end )
|> Stream.drop_while( fn i -> bool = (i<c); side_effect(bool, "c_#{i}"); bool end )
|> Enum.to_list
end
end
When this runs (Demo.go) I get:
a_2
a_3
b_4
b_5
b_7
c_8
c_9
c_10
c_11
c_12
[]
just as I wished - getting a side effect executed for each element in the input list, with an empty list as final output.
But is it possible to generalise this so I can programmatically include (based on a list) as many lines as I like such a:
|> Stream.drop_while( fn i -> bool = (i<x); side_effect(bool, "x_#{i}"); bool end )
I was hoping not to explore macros if I can help it.

Answer 1

A stream is a data structure, which means you can reduce over it, refining it to a particular set of values on every step:
defmodule Demo do
defp side_effect(var, threshold, label) do
if var < threshold do
IO.puts "#{label}_#{var}"
true
else
false
end
end
def go do
values = [a: 4, b: 8, c: 13]
stream =
Enum.reduce(values, [2,3,4,5,7,8,9,10,11,12], fn {k, v}, acc ->
Stream.drop_while(acc, fn i -> side_effect(i, v, k) end)
end)
Enum.to_list(stream)
end
end
You can also explore other solutions. For example, instead of creating a stream for each part of value, you can simple do a filter operation that checks if the value is less than a threshold in a list. Something like this:
defmodule Demo do
defp side_effect(i, values) do
pair = Enum.find(values, fn {_, v} -> i < v end)
case pair do
{k, v} ->
IO.puts "#{k}_#{i}"
false
nil ->
true
end
end
def go do
values = [a: 4, b: 8, c: 13]
[2,3,4,5,7,8,9,10,11,12]
|> Stream.filter(fn i -> side_effect(i, values) end)
|> Enum.to_list()
end
end
I am not sure if you actually need to filter. If not, Stream.map/2 or Stream.each/2 (which is exclusive for side-effects) would work better.