recordclass

What it is and what it does

Recordclass is a library for creating lightweight, mutable record types in Python that avoid the memory overhead of standard dataclasses and namedtuples by opting out of cyclic garbage collection. It provides three main interfaces: a `recordclass` factory (namedtuple-like), a `dataobject` base class (dataclass-like), and lightweight container types (`litetuple`, `mutabletuple`) designed for scenarios where you are certain your objects will never participate in reference cycles.

The library is built on a custom metaclass that skips the `PyGC_Head` prefix normally added to Python objects, reducing per-instance memory by 16–32 bytes depending on Python version. This is most useful for applications that create many small, short-lived objects with simple field types (int, float, str, date/time), such as coordinate records, scientific data points, or compact data structures in memory-constrained environments. The trade-off is that you must manually ensure no circular references are created; the library does not enforce this at runtime.

Use it for

• Store millions of coordinate or vector records (x, y, z) in memory-intensive scientific or graphics applications.
• Build compact data structures for machine learning feature vectors where memory footprint directly impacts cache performance.
• Replace namedtuple in high-throughput data pipelines where object creation speed and size matter.
• Create immutable or mutable tuple-like containers (litetuple, mutabletuple) for collections that will never form cycles.
• Define strongly-typed record classes with type hints that consume less memory than standard dataclass instances.

Worth the install?

AI-flagged interpretation of the facts on this page. Verify before relying on it.

Install

recordclass on PyPI

Before you install

Installation requires compilation from source (high friction). The package is actively maintained with recent commits and supports modern Python versions (3.8–3.14), but the build dependency may complicate deployment in restricted environments.

Requires a C compiler and Python development headers to build from source; pre-built wheels may not be available for all platforms.

License in practice

MIT License permits unrestricted use, modification, and distribution with minimal restrictions—suitable for both open-source and proprietary projects.

Quickstart

from recordclass import recordclass, dataobject, astuple, asdict

# Factory function
Point = recordclass('Point', 'x y')
p = Point(1, 2)
p.y = -1
print(astuple(p))  # (1, -1)
print(asdict(p))   # {'x': 1, 'y': -1}

# Class-based
class Coord(dataobject):
    x: int
    y: int

c = Coord(3, 4)
print(c.x, c.y)

Package facts

Evidence: recordclass-0.24.tar.gz

Tags

More Python Modules packages

See also namedlist · recordtype · dataclass-csv · tinsel · dataclass-factory · HeapDict · ordered-set · itemadapter · fields · dataclasses