Guide: custom overrides

The default output is realistic, but sometimes you need a specific field to hold specific values — a curated set of product names, a fixed status, a foreign key drawn from a particular queryset. Overrides let you pin only those fields and let the engine generate everything else.

Overrides are a Python-API feature (seed(..., overrides=...)).

Override shapes

You pass	The engine does
a value	uses it as-is for every row
a callable	calls it per row for a fresh value
a range	samples an item from it
a list/tuple	samples an item from it
a QuerySet (for an FK)	uses it as a reuse pool, resolved once
a model instance / pk (for an FK)	points every row at it

Example

import random
from shop.models import Book, Author

REAL_TITLES = ["The Pragmatic Programmer", "Clean Code", "Refactoring"]

seed(Book, count=50, overrides={
    "title": lambda: random.choice(REAL_TITLES),  # curated names
    "price": range(5, 100),                        # sampled 5–99
    "status": "published",                         # constant
    "author": Author.objects.filter(active=True),  # FK reuse pool
})

Everything not listed — slug, ISBN, dates, description, other FKs — is still generated and constraint-checked as usual.

Single model vs several

For a single target model, overrides is a {field_name: override} map (as above). When you seed several models at once, key the overrides by model label:

seed(["shop.Book", "shop.Author"], count=20, overrides={
    "shop.Book": {"status": "published"},
    "shop.Author": {"is_active": True},
})

Overrides and uniqueness

Overrides on a unique column still go through the uniqueness tracker, so a sampled/generated override won't produce duplicates. If your override pool is smaller than count for a unique column, generation will exhaust — give it a pool at least as large as the row count.