Python Memory Model, Object References & Mutable Defaults

In Python, everything is an object, and variables are merely labels (references) pointing to boxed PyObject structs in memory.

1. Id, Type, and Value

• id(x): Memory address of the object in CPython.
• is vs ==: == checks value equality; is checks memory identity (id(a) == id(b)).

2. The Mutable Default Argument Bug

• The default list [] is evaluated once at function definition time, NOT at invocation! Every call without tags appends to the exact same list instance!
• Fix: Use tags=None and initialize inside function: if tags is None: tags = [].

2026 Python Internals for Data Pipelines

• Everything Is a PyObject Pointer: Variables bind names to heap objects with reference counts (sys.getrefcount). Assigning b = a on a dict/list aliases the exact same mutable object—always use .copy() or copy.deepcopy() when mutating pipeline state configs!
• Mutable Default Trap: Defining def append_record(row, buf=[]): creates one shared list once at module import time across all future Airflow/ETL runs. Always default to buf=None and initialize if buf is None: buf = [].
• Memory Overhead vs Typed Slots: A plain Python dict record of 5 fields costs ~232+ bytes; __slots__ dataclasses slash per-record RAM by 50%+ when transforming millions of domain events before PyArrow/Parquet serialization.