Elixir Thinking

Mental shifts required before writing Elixir. These contradict conventional OOP patterns.

The Iron Law

NO PROCESS WITHOUT A RUNTIME REASON

Before creating a GenServer, Agent, or any process, answer YES to at least one:

1. Do I need mutable state persisting across calls?
2. Do I need concurrent execution?
3. Do I need fault isolation?

All three are NO? Use plain functions. Modules organize code; processes manage runtime.

The Three Decoupled Dimensions

OOP couples behavior, state, and mutability together. Elixir decouples them:

OOP Dimension	Elixir Equivalent
Behavior	Modules (functions)
State	Data (structs, maps)
Mutability	Processes (GenServer)

Pick only what you need. "I only need data and functions" = no process needed.

"Let It Crash" = "Let It Heal"

The misconception: Write careless code. The truth: Supervisors START processes.

• Handle expected errors explicitly ({:ok, _} / {:error, _})
• Let unexpected errors crash → supervisor restarts

Control Flow

Pattern matching first:

• Match on function heads instead of if/else or case in bodies
• %{} matches ANY map—use map_size(map) == 0 guard for empty maps
• Avoid nested case—refactor to single case, with, or separate functions

Error handling:

• Use {:ok, result} / {:error, reason} for operations that can fail
• Avoid raising exceptions for control flow
• Use with for chaining {:ok, _} / {:error, _} operations

Be explicit about expected cases:

• Avoid _ -> nil catch-alls—they silently swallow unexpected cases
• Avoid value && value.field nil-punning—obscures actual return types
• When a case has {:ok, nil} -> nil alongside {:ok, value} -> value.field, use with instead:

# Verbose
case get_run(id) do
  {:ok, nil} -> nil
  {:ok, run} -> run.recommendations
end

# Prefer
with {:ok, %{recommendations: recs}} <- get_run(id), do: recs

Polymorphism

For Polymorphism Over...	Use	Contract
Modules	Behaviors	Upfront callbacks
Data	Protocols	Upfront implementations
Processes	Message passing	Implicit (send/receive)

Behaviors = default for module polymorphism (very cheap at runtime) Protocols = only when composing data types, especially built-ins Message passing = only when stateful by design (IO, file handles)

Use the simplest abstraction: pattern matching → anonymous functions → behaviors → protocols → message passing. Each step adds complexity.

When justified: Library extensibility, multiple implementations, test swapping. When to stay coupled: Internal module, single implementation, pattern matching handles all cases.

Data Modeling Replaces Class Hierarchies

OOP: Complex class hierarchy + visitor pattern. Elixir: Model as data + pattern matching + recursion.

{:sequence, {:literal, "rain"}, {:repeat, {:alternation, "dogs", "cats"}}}

def interpret({:literal, text}, input), do: ...
def interpret({:sequence, left, right}, input), do: ...
def interpret({:repeat, pattern}, input), do: ...

Defaults and Options

Use /3 variants (Keyword.get/3, Map.get/3) instead of case statements branching on nil:

# WRONG
case Keyword.get(opts, :chunker) do
  nil -> chunker()
  config -> parse_chunker_config(config)
end

# RIGHT
Keyword.get(opts, :chunker, :default) |> parse_chunker_config()

Don't create helper functions to merge config defaults. Inline the fallback:

# WRONG
defp merge_defaults(opts), do: Keyword.merge([repo: Application.get_env(:app, :repo)], opts)

# RIGHT
def some_function(opts) do
  repo = opts[:repo] || Application.get_env(:app, :repo)
end

Idioms

• Process dictionary is typically unidiomatic—pass state explicitly
• Reserve is_thing names for guards only
• Use structs over maps when shape is known: defstruct [:name, :age]
• Prepend to lists [new | list] not list ++ [new]
• Use dbg/1 for debugging—prints formatted value with context
• Use built-in JSON module (Elixir 1.18+) instead of Jason

Testing

Always prefix mix commands with unbuffer to get ANSI colors and prevent stdout block-buffering in non-TTY environments (e.g. unbuffer mix test). Install: brew install expect (macOS) or apt install expect (Linux).

Prefer pattern matching over imperative assertions. Never use assert length + Enum.at/List.last/hd. Pattern match checks length and content in one shot:

# Bad
assert length(students) == 2
assert Enum.at(students, 0).name == "Alice"
assert Enum.at(students, 1).name == "Bob"

# Good
assert [%{name: "Alice"}, %{name: "Bob"}] = students

Same goes for type-only predicates: assert is_map(user) / assert is_list(posts) pass for almost any non-error return. Pattern match the shape and content together: assert %User{email: "a@b.com"} = user. is_nil/1 is fine when nil-ness is the whole point.

Test behavior, not implementation. Test use cases / public API. Refactoring shouldn't break tests.

Test your code, not the framework. If deleting your code doesn't fail the test, it's tautological.

Keep tests async. async: false means you've coupled to global state. Fix the coupling:

Problem	Solution
Application.put_env	Pass config as function argument
Feature flags	Inject via process dictionary or context
ETS tables	Create per-test tables with unique names
External APIs	Use Mox with explicit allowances
File system operations	Use @tag :tmp_dir (see below)

Use tmp_dir for file tests. ExUnit creates unique temp directories per test, async-safe:

@tag :tmp_dir
test "writes file", %{tmp_dir: tmp_dir} do
  path = Path.join(tmp_dir, "test.txt")
  File.write!(path, "content")
  assert File.read!(path) == "content"
end

Directory is auto-cleaned before each run. Works with @moduletag :tmp_dir for all tests in module.

Common Rationalizations

Excuse	Reality
"I need a process to organize this code"	Modules organize code. Processes are for runtime.
"GenServer is the Elixir way"	Plain functions are also the Elixir way.
"I'll need state eventually"	YAGNI. Add process when you need it.
"It's just a simple wrapper process"	Simple wrappers become bottlenecks.
"This is how I'd structure it in OOP"	Rethink from data flow.

Red Flags - STOP and Reconsider

• Creating process without answering the three questions
• Using GenServer for stateless operations
• Wrapping a library in a process "for safety"
• One process per entity without runtime justification
• Reaching for protocols when pattern matching works

Any of these? Re-read The Iron Law.