introduction

erlang

• ericsson language
• joe armstrong
• 31 years ago
• beam vm

goals

• scalability
• distribution
• fault tolerance
• high availability
• hot-code swapping 

who uses erlang?

basically everyone already used it in some way

whatsapp

• messaging servers
• 2011: 1 million concurrent tcp connections
• 2012: they reached 2 million

The WhatsApp architecture Facebook bought for 19 billion

* http://highscalability.com/blog/2014/2/26/the-whatsapp-architecture-facebook-bought-for-19-billion.html

uses erlang to implement its chat service, handling more than 2 billion montly active users

uses erlang to implement simpledb, a database service of the amazon elastic compute cloud

uses erlang in its sms and authentication systems

migrated from ruby to elixir and went from 150 servers to just 5

erlang?

isn't this talk about elixir?

elixir=erlang

• jose valim
• compiles to erlang
• ruby-like syntax (jose was part of the rails core team)
• 1.0 released 2011 after 8005 commits
• it is functional! (more about this later)

By the way

Ask questions any time!

functional vs oop

a historical perspective*

• “I guess by end of 70's people were burnt out [...] it made sense that what you can understand better will be manageable better.”
• “Object-oriented methodology with a promise '...everything in life is an object' seemed more like common sense even before it was proven to be meaningful.”
• “[...] this paradigm forced you to somewhat think before you code more than procedural languages, it showed visible success on the software projects that employed, and since then they caught on!”

* https://softwareengineering.stackexchange.com/questions/137716/what-were-the-historical-conditions-that-led-to-object-oriented-programming-beco

the biggest differences*

object-oriented programming:

• a programmer creates objects
• encapsulation: data and operations are kept together
• inheritance: classes can be reused
  • more and more criticized because if used intensively, it hinders reuse and agility
  • composition is an alternative
• polymorphism: term used for a lot of things, interfaces are an alternative
  • function overloading
  • generics
  • subtype polymorphism

* https://medium.com/@cscalfani/goodbye-object-oriented-programming-a59cda4c0e53 and https://stackoverflow.com/questions/1112773/what-are-the-core-concepts-in-functional-programming

the biggest differences*

functional programming:

• a programmer creates functions
• data and operations are kept separate
• no inheritance, composition
• polymorphism
  • “function overloading” (guards)
  • “generics”
  • “interfaces” (protocols)

* https://medium.com/@cscalfani/goodbye-object-oriented-programming-a59cda4c0e53 and https://stackoverflow.com/questions/1112773/what-are-the-core-concepts-in-functional-programming

an intuitive explanation

one thing they sometimes teach you in programming classes is how to study requirements to extract classes and methods.

1. "The user shall be able to search either all of the initial set of databases or select a subset from it."
2. "The system shall provide appropriate viewers for the user to read documents in the document store."
3. "Every order shall be allocated a unique identifier (ORDER_ID) which the user shall be able to copy to the account’s permanent storage area."

is functional/oop programming a hype?*

* https://www.quora.com/Is-the-object-oriented-programming-OOP-paradigm-overemphasized-or-overrated

60ies

90ies

what oop users claim*

* http://i.imgur.com/Q0vFcHd.png

what actually happens*

* http://i.imgur.com/Q0vFcHd.png

what to choose?

• Object-oriented languages are good when you have a fixed set of operations on things, and as your code evolves, you primarily add new things. This can be accomplished by adding new classes which implement existing methods, and the existing classes are left alone.
• Functional languages are good when you have a fixed set of things, and as your code evolves, you primarily add new operations on existing things. This can be accomplished by adding new functions which compute with existing data types, and the existing functions are left alone.

[5] https://stackoverflow.com/questions/2078978/functional-programming-vs-object-oriented-programming

the tooling system

two tools to rule them all...

mix

mix help new

project structure

iex

the interactive shell

pattern matching

a = 1 # 1

1 = a # 1 - also a valid statement

2 = a # MatchError

= does not mean to assign, it is a match operator

{:ok, message} = {:ok, "hello, elixir"}
message # "hello, elixir"

{:status, code} = {:status, 404}
code # 404

{:error, err, _} = {:error, "some error message"} # MatchError - 
# tuples with different sizes

{:ok, :hi} = {:ok, "hi"} # MatchError

tuples

list = [1, 2, "last"]
[head|tail] = list
head # 1
tail # [2, "last"]

lists

can also be used to prepend items to a list

list = [0|list]
list # [0, 1, 2, "last"]

list = [{:a, 1}, {:b, 2}] # - can also be written [a: 1, b: 2]

[a: x, b: y] = list # - same number of tuples to match
x # 1
y # 2

keyword lists

map = %{:name => "jose", :surname => "valim"} # elixir's creator

%{name: "jose", surname: "valim"} == map # true - convinient way to write
# maps when keys are atoms only

%{"name" => x} = map
x # "jose"

maps

a = 1
a # 1
{a, b} = {"foo", 2}
a # "foo" - ...

the pin ^ operator

how to match something without hardcoding?

using ^

a = 1
a # 1
{^a, b} = {"foo", 2} # MatchError - hurray
{^a, b} = {1, 2} # - ^ does not allow variable rebind

the pipe |> operator

|>

foo(bar(baz(new_function(other_function()))))

instead of writing

if? what?

control flow

the elixir way

pattern matching, pattern matching, pattern matching

control flow (1)

# match and bind `code`
def handle_response(:ok, %{status: code}) do

# match and bind `message`
def handle_response(:error, message) do

# match everything and bind data with a guard
def handle_response(_, data) when is_map(data) do

# match everything and bind data
def handle_response(_, data) do

parameters

control flow (2)

with {:ok, url} <- build_url(),
     {:ok, response} <- fetch_users(url),
     {:ok, users} <- parse_response(response) do
  
  do_something(users)
else
  # error handling
end

with

looping

functional languages

do not need for loops

recursion

# elixir

def sum([]), do: 0
def sum([h|t]), do: h + sum(t)

// golang

func sum(l []int) int {
	result := 0
	for _, v := range l {
		result += v
	}

	return result
}

example: summing integers in a list

# result

sum([1, 2, 3, 4]) = 10

tail recursion

# recursion

def sum([]), do: 0
def sum([h|t]), do: h + sum(t)

# tail recursion

def sum(l), do: do_sum(l, 0)

defp do_sum([], acc), do: acc
defp do_sum([h|t], acc), do: do_sum(t, acc + h)

odd? = &(rem(&1, 2) != 0)

Enum and Stream

the std library provides also a lot of functions for handling with enumerables (such as list and map)

(intro to) processes

spawn

self() # #PID<0.84.0> - what is this?

sending messages

let's modify our function a bit

receiving messages

f = fn pid -> send(pid, {:mess, "hello world, from #{inspect self()}"}) end

let's start again with our function

defmodule Counter do    
    def start(val), do: spawn(__MODULE__, :loop, [val])

    def loop(state) do
        receive do
            {:status, pid} ->
                send(pid, "status -> #{inspect state}")
                loop(state)
            :incr ->
                loop(state + 1)
            {:stop, pid} ->
                send(pid, "stopping. last status -> #{inspect state}")
        end
    end
end

common pattern

iex(3)> pid = Counter.start(0)
iex(4)> send(pid, {:status, self()})
iex(5)> flush
"status -> 0"
iex(6)> send(pid, :incr)
iex(7)> send(pid, {:status, self()})
iex(8)> flush
"status -> 1"

links (1)

# raise/1 causes a RuntimeError
f = fn -> raise("ouch") end

self() # #PID<0.120.0>

pid = spawn_link(f)
# ** (EXIT from #PID<0.104.0>) evaluator process exited with reason: 
# an exception was raised:
# ** (RuntimeError) ouch
#     (stdlib) erl_eval.erl:668: :erl_eval.do_apply/6
#
# 15:29:04.053 [error] Process #PID<0.110.0> raised an exception
# ** (RuntimeError) ouch
# (stdlib) erl_eval.erl:668: :erl_eval.do_apply/6

Process.alive?(pid) # false - crashed immediately, as expected

self() # #PID<0.126.0> - we crashed as well!

links (2)

making a process crash using spawn_link

links (3)

f = fn -> receive do :crash -> raise "ouch" end end

making a process crash using spawn_link and trapping exit signals

no boundaries

send(dest, message) # - from the official documentation
...
dest may be a remote or local PID, a (local) port, a locally registered name, 
or a tuple {registered_name, node} for a registered name at another node.

defensive programming

Have you ever programmed something like this?

URL url = null;
try {
	url = new URL(urlAsString);
} catch (MalformedURLException e) {
	logger.error("Malformed URL", e);
}
 
HttpURLConnection connection = null;
try {
	connection = (HttpURLConnection) url.openConnection();
} catch (IOException e) {
	logger.error("Could not connect to " + url, e);
}

StringBuilder builder = new StringBuilder();
try (BufferedReader reader = new BufferedReader(new ...

"let it crash"

“Let it crash” refers to a coding philosophy that “if there are any errors, the process is automatically terminated, and this is reported to any processes that were monitoring the crashed process.”