django-picklefield

django-picklefield provides PickledObjectField, a Django ORM model field that stores any picklable Python object — lists, dicts, custom classes, nested structures — directly on a model without needing to design a relational schema or JSON-serializable representation for it. Assigning a value to the field behaves like any other model attribute; the pickling and unpickling happen transparently on save/load.

Originally adapted from community Django snippets, it has become a common utility for storing loosely structured or legacy Python data inside a Django model column when a proper relational schema isn’t worth the design cost, such as caching computed results or persisting arbitrary configuration blobs.

What You Get

• PickledObjectField — a drop-in Django model field accepting any picklable Python object
• Transparent serialization/deserialization on model save/load with no manual encode/decode calls
• Support for default, null, editable, and other standard Django field options
• Compression support for large pickled payloads to reduce storage size
• Compatibility layer constants (picklefield/constants.py) for handling protocol/version differences across Python and Django releases

Common Use Cases

• Caching computed Python objects (e.g. ML model outputs, parsed structures) directly on a Django model row
• Persisting loosely structured configuration or metadata that doesn’t warrant a dedicated relational schema
• Migrating legacy Django codebases that already relied on pickled blobs in the database
• Storing arbitrary Python objects (not just JSON-safe primitives) associated with a model instance

Under The Hood

Architecture: The package is a focused Django field extension: picklefield/fields.py (221 lines) implements PickledObjectField by subclassing Django’s field base classes and overriding the value-to/from-database conversion hooks to pickle and unpickle transparently, while picklefield/constants.py centralizes protocol and compatibility constants.

Tech Stack: Pure Python built directly on Django’s ORM field API and the standard library pickle module, with no external runtime dependencies beyond Django itself.

Code Quality: A single consolidated test module (tests/tests.py) exercises field behavior against a real Django test app, with CI configured via GitHub Actions (test.yml) and coverage tracked via Coveralls per the README badges.

API Design: Usage is a one-line addition to a model — args = PickledObjectField() — mirroring any other Django field, so there’s essentially no learning curve for anyone already familiar with Django models; the tradeoff is inherited from pickle itself (security and cross-version compatibility caveats), which the library documents rather than abstracts away.