Count words in string and tally up the repetitive words

Question

I'm an elixir beginner. I'm doing the word count exercise of exercism website. The word count exercise returns a map with the word as the key and the number of instances as value. The exercise excludes special characters, and ignores underscores. My code works, but I'm sure there is a way to make it cleaner. I've attached the test file so that you can see the criteria that it needs to pass.

word_count.exs

defmodule Words do
 @doc """
 Count the number of words in the sentence.
 Words are compared case-insensitively.
 """
 @spec count(String.t) :: map
 def count(sentence) do
 String.downcase(sentence) |>
 String.split(~r/[\s_]/) |>
 Enum.filter(fn(word) -> String.match?(word, ~r/[a-zA-z\d]/) end) |>
 Enum.map(fn(word) -> String.replace(word, ~r/[:!&@$%^,]/, "") end)|>
 Enum.group_by(fn(word) -> word end) |>
 Enum.reduce(%{}, fn({k, v}, acc) -> Map.put(acc, k, Enum.count(v)) end)
 end
end

word_count_test.exs

if !System.get_env("EXERCISM_TEST_EXAMPLES") do
 Code.load_file("word_count.exs")
end
ExUnit.start
ExUnit.configure exclude: :pending, trace: true
defmodule WordsTest do
 use ExUnit.Case
 test "count one word" do
 assert Words.count("word") == %{ "word" => 1 }
 end
 test "count one of each" do
 expected = %{ "one" => 1 , "of" => 1 , "each" => 1 }
 assert Words.count("one of each") == expected
 end
 test "count multiple occurrences" do
 expected = %{ "one" => 1 , "fish" => 4 , "two" => 1 , "red" => 1 , "blue" => 1 }
 assert Words.count("one fish two fish red fish blue fish") == expected
 end
 test "ignore punctuation" do
 expected = %{"car" => 1, "carpet" => 1, "as" => 1, "java" => 1, "javascript" => 1}
 assert Words.count("car : carpet as java : javascript!!&@$%^&") == expected
 end
 test "include numbers" do
 expected = %{"testing" => 2, "1" => 1, "2" => 1}
 assert Words.count("testing, 1, 2 testing") == expected
 end
 test "hyphens" do
 expected = %{"co-operative" => 1}
 assert Words.count("co-operative") == expected
 end
 test "ignore underscores" do
 expected = %{"two" => 1, "words" => 1}
 assert Words.count("two_words") == expected
 end
 test "normalize case" do
 expected = %{"go" => 3}
 assert Words.count("go Go GO") == expected
 end
 test "German" do
 expected = %{"götterfunken" => 1, "schöner" => 1, "freude" => 1}
 assert Words.count("Freude schöner Götterfunken") == expected
 end
end

Answer 1

Just finished that exercise too. I think you can extract some of that logic to anonymous functions. It isolates a pieces of behaviour that you can work on.

defmodule Words do
 @doc """
 Count the number of words in the sentence.
 Words are compared case-insensitively.
 """
 @spec count(String.t) :: map
 def count(sentence) do
 filter_word = fn(word) ->
 String.match?(word, ~r/[a-zA-z\d]/)
 end
 remove_special_characters = fn(word) ->
 String.replace(word, ~r/[:!&@$%^,]/, "")
 end
 count_words = fn({k, v}, acc) ->
 Map.put(acc, k, Enum.count(v))
 end
 sentence
 |> String.downcase
 |> String.split(~r/[\s_]/)
 |> Enum.filter(filter_word)
 |> Enum.map(remove_special_characters)
 |> Enum.group_by(fn(word) -> word end)
 |> Enum.reduce(%{}, count_words)
 end
end

And actually, I've created a function that returns a list of words to count: http://exercism.io/submissions/e26667aff3734da79ac0b6c6fedfd3da

Hope that helps!