zspec 1.4.0

Composable specification pattern for Python

zspec

Composable specification pattern for Python 3.14+.

Install

pip install zspec

Quick start

from dataclasses import dataclass
from zspec import Specification


@dataclass
class Product:
    price: int
    in_stock: bool


# Define rules as classes
class InStock(Specification[Product]):
    def is_satisfied_by(self, p: Product) -> bool:
        return p.in_stock


class MinPrice(Specification[Product]):
    def __init__(self, threshold: int) -> None:
        self.threshold = threshold

    def is_satisfied_by(self, p: Product) -> bool:
        return p.price >= self.threshold


# Compose with &, |, ^, ~
eligible = InStock() & MinPrice(100)

product = Product(price=200, in_stock=True)
assert eligible(product)

Or skip the class boilerplate:

# Field comparisons
spec = Specification[Product].matching(price__gte=100, in_stock=True)

# Lambda predicates
spec = Specification[Product].matching(
    lambda p: p.price > 100,
    lambda p: p.in_stock,
)

Why zspec?

Composable	& | ^ ~ — build complex rules from simple ones without new classes
Zero dependencies	Standard library only. Optional extras for SQLAlchemy, Django, and Polars
Type-safe	Generic Specification[T] preserves candidate types through composition
Database translators	One spec → SQL, MongoDB, Django Q, SQLAlchemy, or Polars expression
Serializable	to_dict() / from_dict() — store rules in JSON configs or databases
Debuggable	explain() prints a PASS / FAIL tree for every node

Filtering collections

passed = list(eligible.filter(products))   # lazy generator
failed = list(eligible.reject(products))   # inverse
passed, failed = eligible.partition(products)
count = eligible.count(products)

Debugging

from zspec import explain

print(explain(eligible, product))
# AND FAIL
# ├── InStock PASS
# └── price >= 100 FAIL

Serialization

from zspec import to_dict, from_dict, registered

@registered
class InStock(Specification[Product]):
    ...

# Save
json.dump(to_dict(InStock() & MinPrice(100)), f)

# Load — @registered specs are auto-discovered
spec = from_dict(json.load(f))

Translators

One spec — query any backend:

# SQL
MySql().translate(eligible)          # SqlFragment("price >= %s AND in_stock", (100, True))

# MongoDB
MyMongo().translate(eligible)        # {"$and": [{...}, {...}]}

# Django
MyDjango().translate(eligible)       # Q(price__gte=100) & Q(in_stock=True)

# SQLAlchemy
MySA().translate(eligible)           # ColumnElement[bool]

# Polars
MyPolars().translate(eligible)       # pl.Expr

License

MIT