494: Primitive Obsession

What It Covers

Joelle and Sally explore primitive obsession in software development — the practice of overusing raw types like strings, integers, and arrays instead of domain-specific objects. They also cover graph data structure storage in Postgres, including adjacency lists, recursive CTEs, closure tables, and materialization strategies for representing contractor relationships.

Key Questions Answered

• •Primitive Obsession Definition: Primitive obsession occurs when programmers project implicit semantics onto low-level types. An array used as a shopping cart forces every call site to re-implement shopping cart logic — subtotal calculations, tax handling — rather than centralizing that behavior in a dedicated object with encapsulated methods. This scatters domain logic across controllers, views, and services.
• •Type Systems as Bug Prevention: In typed languages like Elm, creating domain-specific types for measurements prevents argument-order bugs. A function accepting an hourly `Rate` type and an `Hours` type will reject swapped arguments at compile time, whereas two plain integers are indistinguishable. Ruby keyword arguments offer partial protection, but dedicated types provide stronger guarantees.
• •Null Object Pattern as Primitive Obsession: Using `nil` to represent a guest user is a form of primitive obsession — nil means absence of data, not "guest." When nil carries context-specific meaning that callers must implicitly know, a dedicated `GuestUser` object eliminates that hidden knowledge and removes conditional nil-checks scattered throughout the codebase.
• •Graph Storage Strategy — Start Simple: For storing graph relationships in Postgres, begin with an adjacency list table as the single source of truth, then query it using recursive CTEs. Only introduce a closure table (denormalized, easier to query) when performance demands it, treating it as a materialized derivative that can be truncated and regenerated from the adjacency list at any time.
• •Refactoring Away from Primitives — Use Structs First: When resisting the creation of a full class feels like too large a step, start with a Ruby `Struct` or `Data` class inline. This establishes a named interface immediately and can be promoted to a full class once behavior accumulates. Large `map` or `reduce` blocks with multi-line logic are a reliable code smell signaling a missing domain object.