Engineering Craft › Clean Code & Principles

Primitive Obsession

Using raw strings and ints where a small domain type belongs.

Also known as: primitive obsession smell, stringly typed

Primitive obsession is the habit of representing domain ideas with basic types such as strings, integers and floats. An email address is a string, a price is a float, and an order status is a string that happens to be one of three words. The types don’t stop nonsense from getting in, and the rules about each idea end up scattered across the code.

# Primitives: any string and any number is accepted
def send_receipt(email: str, amount: float, currency: str): ...

send_receipt("not-an-email", -5, "dollars")   # runs without complaint

A small domain type makes the rules live in one place and the mistakes visible:

from dataclasses import dataclass

@dataclass(frozen=True)
class Money:
    cents: int
    currency: str

    def __post_init__(self):
        if self.cents < 0:
            raise ValueError("amount must not be negative")

The trade-off is extra types. Each new class is a file, a name and a conversion point, and small projects can be buried in wrappers. Not every string needs a type: a free-text note or a log message is fine as a plain string.

The classic mistake is validating the same string in five places, each slightly differently, because nothing marks it as an email or a currency code. Introduce a type for ideas that have rules, that get passed around, or that are easy to mix up, such as two IDs of different kinds. Once the rules sit in one value object, shotgun surgery becomes much less likely when they change.