sqls 0.0.100

SQL Toolkit for Python.

SQLs

SQLs is a collection of libraries to interact with SQL databases. SQLs is not an object-relational mapper (ORM), but offers useful low-level primitives for handling transactions, defining a data model, and formulating queries in Python. Its main design goal is helping developers write SQL in idiomatic, type-checked Python, while being fast and efficient on runtime.

Getting started

For more, see the documentation.

SQLs is published on PyPI and does not have any special dependencies. Simply install with pip:

pip install sqls

Dependencies on database interface libraries are strictly optional. If you want to speak to a specific SQL implementation use any of:

pip install sqls[mysql]
pip install sqls[postgresql]
pip install sqls[sqlite]
pip install sqls[mysql,postgre,sqlite]

Connect to a database

Asynchronous transaction managers from sqls.transactions handle SQL database connections:

from sqls.transactions import get_manager

async def application() -> list[tuple[int]]:
    # Create a transaction manager.
    manager = get_manager('file:///path/to/sqlite.db')

    # Initialize database connections.
    await manager.init()

    # Open transaction.
    async with manager.txn() as txn:

       # Execute query.
       return await txn.execute('SELECT 1')

    # Close database connections.
    await manager.close()

All SQL statements inside the asynchronous context manager are executed in one single transaction. Uncaught exceptions in the context will cause the transaction to be automatically rolled back, on regular exit the transaction will automatically be commited.

Define a data model

The data model is defined through annotated Python dataclasses and the Model base class from sqls.models.

Basic usage

The syntax uses builtin Python primitives to express the rich details of SQL types:

from dataclasses import dataclass
from typing import Annotated

from sqls.models import CharField, Fieldmeta, IntegerField, Model

@dataclass
class User(Model):
    """A user model."""

    # Names are unique.
    name: Annotated[CharField, Fieldmeta(max_length=32, unique=True)]

    # Passwords are nullable.
    password: Annotated[CharField, Fieldmeta(max_length=128)] | None

    # Use just a plain integer.
    time_created: IntegerField

The Model base class automatically adds an integer primary key id field.

Relationships

Relationships are expressed through annotations on fields that store the actual information:

@dataclass
class User(Model):
    """Same as above, add some relationships."""

    # Table name and field are inferred from the attribute name.
    company_id: ForeignKeyField

    # Table name and field are explicitly set though Fieldmeta.
    team_id: Annotated[ForeignKeyField, Fieldmeta(foreign_key=('department', 'id'))

Many-to-many relationships cannot be expressed on the related models themselves, the table needs to be defined explicitly:

@dataclass
class UserGroup(Model):
    """User group relationship."""

    # Disable automatic injection of id field.
    id: None
    user_id: ForeignKeyField
    group_id: ForeignKeyField

Create database tables

The sqls.models package can generate CREATE TABLE queries from model definitions:

from sqls.models import get_create_queries

# Inside a transaction context (txn) create tables for User and Group.
for query in get_create_queries([User, UserGroup, Group]):
    # Execute generated query with txn.exq.
    await txn.exq(query)

Build queries

The sqls.queries package helps writing queries in idiomatic python:

from sqls.queries import Query, Table

# Create Table object from sql table name.
user_t = Table('user')

# Create query for id and password of one specific user.
query = (
    Query
    .select(
        user_t.id.typed(int),
        user_t.password.typed(str | None),
     )
    .from_(user_t)
    .where(user_t.name == 'Ringo')
)

As SQLs is not an ORM, Query knows nothing about the data model. By expressing the expected return type of the id field with .typed(int) static typed checkers like mypy are able to infer the return types when the query is executed.

Development

Clone the repository and setup your local checkout:

git clone https://gitlab.com/durko/sqls.git

cd sqls
python -m venv venv
. venv/bin/activate

pip install -r requirements-dev.txt
pip install -e .

This creates a new virtual environment with the necessary python dependencies and installs SQLs in editable mode. The SQLs code base uses pytest as its test runner, run the test suite by simply invoking:

pytest

To build the documentation from its source run sphinx-build:

sphinx-build -a docs build/public

The entry point to the local documentation build should be available under build/public/index.html.