One of the points of yield/generators in languages like ruby and python is to compute solutions incrementally instead of materializing the solution all at once thus allowing large or even infinite results to be computed. While Elixir doesn’t have yield, it achieves the same goal using Streams. Let me illustrate with a simple backtracking problem: n queens. An inefficient but concise formulation in ruby might go:
def solve(n, queens, rank, &block)
if rank == n
yield queens
else
n.times do |file|
if safe?(queens, file, rank)
queens << [file, rank]
solve(n, queens, rank + 1, &block)
queens.pop
end
end
end
end
solve(ARGV[0].to_i, [], 0) { |solution| puts solution.inspect }
The use of yield allows us to use the same code to print one result, all results or count the number of results by supplying a different block to consume then result.
To accomplish the same goal in Elixir, you’d use Streams and flatmap, like so:
def solve(n, queens, r) when r == n, do: [{reverse(queens)}]
def solve(n, queens, r) do
0..(n-1)
|> filter(fn f -> safe?(queens, {f, r}) end)
|> flat_map(fn f -> solve(n, [{f, r} | queens], r+1) end)
end
System.argv() |> at(0) |> String.to_integer |> solve([], 0) |> each(&IO.inspect/1)
Hope that helps.
