Giving up the Object-Oriented Ghost

@morganlaco

github.com/mlaco

hello@morganlaco.com

http://slides.com/morganlaco/deck-2/live

Ghosts

Ghosts

• Objects
• Mutability
• Conditionals
• Loops

What is functional programming?

Functional vs Imperative

Lambda Calculus

Lambda Calculus

2 ≡ λf.λx.f (f x)

+ ≡ λm.λn.λf.λx.m f (n f n)

Lambda Calculus

2 ≡ λf.λx.f (f x)

fn(f) {
  return fn(x) {
    return f(f(x))
}}

Lambda Calculus

2 ≡ λf.λx.f (f x)

fn(f) {
  return fn(x) {
    return f(f(x))
}}

Ghost 1: Objects

The Fours Awakens

github.com/mlaco/ConnectFour

class ConnectFour::Game
  def initialize
    @board = ConnectFour::Board.new
    # other instance variables
  end
  
  def play
    while continue_game
      get_move
      handle_move
    end
    
    display_result
  end
  # ...
end

ConnectFour::Game.new.play

defmodule ConnectFour do
  def start(_type, _args) do
    ConnectFour.initial_state
    |> do_turn
  end
  
  def do_turn(state) do
    ConnectFour.IO.display_board(state[:board])
    
    case ConnectFour.Board.update_board(state[:board], \
        move, state[:turn]) do
      # ... error and quit cases ...
      new_state ->
        if ConnectFour.WinChecker.check(new_state[:board]) do
          ConnectFour.IO.do_win(new_state)
        else  
          do_turn(new_state)
        end
    end
  end

ConnectFour.start

Objects ->

Functional Purity

Ghost 2: Conditionals

  def find_top(col, n\\0) do
    col_list = Tuple.to_list(col)
    [h|t] = col_list
    
    cond do
      !Enum.any?(t) && h != 0 ->
        {:error}
      h == 0 ->
        {:ok, n}
      true ->
        List.to_tuple(t) |> find_top(n+1)
    end
  end

0

1

-1

Key

#

  def find_top(col, n\\0) do
    col_list = Tuple.to_list(col)
    [h|t] = col_list
    
    cond do
      !Enum.any?(t) && h != 0 ->
        {:error}
      h == 0 ->
        {:ok, n}
      true ->
        List.to_tuple(t) |> find_top(n+1)
    end
  end

  def find_top(col, n\\0) do
    col_list = Tuple.to_list(col)
    [h|t] = col_list
    
    cond do
      !Enum.any?(t) && h != 0 ->
        {:error}
      h == 0 ->
        {:ok, n}
      true ->
        List.to_tuple(t) |> find_top(n+1)
    end
  end

h

  def find_top(col, n\\0) do
    col_list = Tuple.to_list(col)
    [h|t] = col_list
    
    cond do
      !Enum.any?(t) && h != 0 ->
        {:error}
      h == 0 ->
        {:ok, n}
      true ->
        List.to_tuple(t) |> find_top(n+1)
    end
  end

h

t

  def find_top(col, n\\0) do
    col_list = Tuple.to_list(col)
    [h|t] = col_list
    
    cond do
      !Enum.any?(t) && h != 0 ->
        {:error}
      h == 0 ->
        {:ok, n}
      true ->
        List.to_tuple(t) |> find_top(n+1)
    end
  end

h

t

Pattern Matching

Pattern Matching

[h|t] = col_list

  def check(remaining, []) do
    [next_row | other_rows] = remaining
    check(other_rows, next_row)
  end
  
  def check([], row) do
    check_row(row)
  end

  def check(remaining, row) do
    if check_row(row) do
      true
    else
      [next_row | other_rows] = remaining
      check(other_rows, next_row)
    end
  end

Inversion of Control

Conditionals ->
Pattern Matching

Ghost 3: Mutability

Mutability

What the fuzz?

immutability

before

after

foo

foo

foo = 4
foo = 5

4

5

immutability

before

after

foo

foo

foo = 4
foo = 5

4

5

4

foo

Why variables can be re-bound

Benefits of Immutability

• Performance
• Distribution

FP in other languages

http://www.lispcast.com/imperative-mindset

Mutability ->
Immutability

Ghost 4: Loops

Loops

(1..100).each do |i|
  do_something(i)
end

while i_am_talking do
  be_nervous
end

No Loops - why tho?

Recursion

  def fibonacci(n) do
    cond do
      n == 0
        0
      n == 1
        1
      true ->
        fibonacci(n, 0, 1)
    end
  end

  def fibonacci(n, a, b) do
    if n <= 0 do
      a
    else
      fibonacci(n - 1, b, a + b)
    end
  end

Recursion

def self.check(grid)
  grid.each_with_index do |column, column_index|
    n = 1
    last = 0
    column.each_with_index do |cell, row_index|
      if cell == last && cell != 0
        n += 1
      else
        n = 1
      end
      last = cell
      
      return true if n == 4
    end
  end

  def check_horizontal(board) do
    Transpose.transpose(board)
    |> check([])
  end

  def check(remaining, []) do
    [next_row | other_rows] = remaining
    check(other_rows, next_row)
  end
  
  def check([], row) do
    check_row(row)
  end

  def check(remaining, row) do
    if check_row(row) do
      true
    else
      [next_row | other_rows] = remaining
      check(other_rows, next_row)
    end
  end

  def check([next_row | other_rows], []) do
    check(other_rows, next_row)
  end
  
  # Equivalent to:
  def check(remaining, []) do
    [next_row | other_rows] = remaining
    check(other_rows, next_row)
  end

  # pass
  def check_row(cell, [], last, matching) do
    new_matching =
      if cell == last do
        matching + 1
      else
        1
      end
    new_matching == 4
  end
  
  # pass
  def check_row(cell, remaining, last, matching) do
    new_matching = 
      if cell == last && cell != 0 do
        matching + 1
      else
        1
      end
    if new_matching == 4 do
      true
    else
      [next_cell | other_cells] = remaining
      check_row(next_cell, other_cells, cell, new_matching)  
    end  
  end

Tail Recursion

tail recursion

how does it work?

fibonacci(n)

fibonacci(n)

fibonacci(n)

fibonacci(n)

<process>

def fibonacci(n) do
  cond do
    n == 0
      0
    n == 1
      1
    true ->
      fibonacci(n, 0, 1)
  end
end

def fibonacci(n, a, b) do
  if n <= 0 do
    a
  else
    fibonacci(n - 1, b, a + b)
  end
end

tail recursion

how does it work?

fibonacci(n)

fibonacci(n)

fibonacci(n)

fibonacci(n)

<process>

def fibonacci(n) do
  cond do
    n == 0
      0
    n == 1
      1
    true ->
      fibonacci(n, 0, 1)
  end
end

def fibonacci(n, a, b) do
  if n <= 0 do
    a
  else
    fibonacci(n - 1, b, a + b)
  end
end

What I told you was true...

Loops

Enum.reduce( (1..100), \
  fn(x, acc) -> x + acc end )
  # => 5050

Enum.map( (1..5), \
  fn(x) -> 2 * x end )
  # => [2, 4, 6, 8, 10]

Enum.filter( (1..5), \
  fn(x) -> rem(x, 2) == 0 end )
  # => [2, 4]

defmodule ConnectFour.WinChecker.OrthogonalHigherOrder do
  def check_horizontal(board) do
    Enum.filter(board, fn(row) -> eval_row(row) end)
    |> Enum.any?(fn(res) -> res end)
  end
  
  # If the row has 4 consecutive, 
  # return [n: 4, prev: c] where c is the winning
  # color
  def eval_row(row) do
    Enum.reduce(row, [n: 0, prev: 0], \
      fn(cell, acc) -> eval_cell(acc, cell) end)
    |> is_there_a_win
  end
  
  def is_there_a_win([n: 4, prev: _]) do
    true
  end
  
  def is_there_a_win(_) do
    false
  end
  
  # If a match has already been found,
  # just pass that along
  def eval_cell([n: 4, prev: a], _) do
    [n: 4, prev: a]
  end
  
  # param
  #  dict acc
  #    n: current number of accumulated consecutive cells
  #    prev: color of previous cell
  # param
  #  cell: current cell color
  def eval_cell(acc, cell) do
    [n: next_n(acc, cell), prev: cell]
  end
  
  # update the number of consecutive same-colored cells 
  def next_n(_, 0) do
    # Reset the count
    0
  end
  
  def next_n([n: n, prev: 0], a) do
    # First of the color
    1
  end
  
  def next_n([n: n, prev: a], a) do
    # Matches with previous
    n + 1
  end
  
  def next_n([n: _, prev: a], b) do
    # No match, reset
    0
  end
end

Loops ->
Recursion

Ghosts Review

references

1. From Imperative to Functional and Back-Monads are for Functional Languages
2. From Imperative to Functional: How to Make the Leap
3. A Practical Introduction to Functional Programming
4. StackOverflow: What are the core concepts in functional programming?
5. Elixir Getting Started Guide
6. Destroy All Ifs
7. 16 Months of Functional Programming
8. Learn You Haskell
9. Is object immutability in functional programming inherently performance intensive?
10. How to Switch from the Imperative Mindset
11. Functional Programming Techniques With Ruby: Part I
12. From Imperative to Functional Programming (for Absolute Beginners)
13. Transposing a Matrix: Thinking Recursively in Elixir
14. Don’t Be Scared Of Functional Programming
15. Myth of the Day: Functional Programmers Don't Use Loops

also check out: elixirschool.com

watch for Elixir Code School courses!

THANK YOU!!!