ysaqml 2025.12.6.post0

ysaqml provides a YAML-backed persistence layer for SQLAlchemy.

ysaqml

Overview

ysaqml provides a YAML-backed persistence layer for SQLAlchemy. Tables live in memory inside SQLite for fast queries, while their contents are loaded from – and saved back to – .yaml files using the blazing-fast naay parser/dumper.

Why?

• Keep relational workflows while still committing plain-text fixtures.
• Edit data in YAML, rely on SQLite for constraints and transactions.
• Round-trip friendly thanks to naay's strict string-only subset.

Installation

pip install -e .

This pulls in SQLAlchemy and naay. Tests rely on pytest (install with pip install -e .[test]).

Usage

Context-managed workflow

from pathlib import Path
from sqlalchemy import Column, Integer, MetaData, String, Table, select

from ysaqml import YamlSqliteEngine

metadata = MetaData()
users = Table(
		"users",
		metadata,
		Column("id", Integer, primary_key=True),
		Column("name", String, nullable=False),
)

data_dir = Path("./data")

with YamlSqliteEngine(metadata, data_dir) as backend:
		engine = backend.engine
		with engine.begin() as conn:
				conn.execute(users.insert().values(id=1, name="Ada"))
				result = conn.execute(select(users)).mappings().all()
				print(result)
# context manager flushes SQLite rows back into data/users.yaml

Each SQLAlchemy table is materialized to <data_dir>/<table_name>.yaml. When entering the context the engine wipes any existing rows in the in-memory database, rehydrates them from YAML if the file exists, and on successful exit dumps the SQLite contents back to YAML.

Custom dialect / engine helper

ysaqml also registers a SQLAlchemy dialect so you can build engines with ysaqml:/// URLs. The convenience helper wires up the required engine plugin and disposes the engine when you are done:

from contextlib import closing
from pathlib import Path
from sqlalchemy import Column, Integer, MetaData, String, Table, select

from ysaqml import create_yaml_engine

metadata = MetaData()
users = Table(
		"users",
		metadata,
		Column("id", Integer, primary_key=True),
		Column("name", String, nullable=False),
)

data_dir = Path("./data")

with closing(create_yaml_engine(metadata, data_dir)) as engine:
		with engine.begin() as conn:
				conn.execute(users.insert().values(id=1, name="Ada"))
				print(conn.execute(select(users)).mappings().all())
# exiting the context disposes the engine and writes YAML files

Under the hood this builds an in-memory SQLite engine using the ysaqml dialect and attaches an engine plugin that loads YAML files immediately and saves them back when the engine is disposed. If you manage the engine manually, make sure to call engine.dispose() to persist your changes.

File format

Every YAML document uses the following schema:

_naay_version: "1.0"
rows:
	- id: "1"
		name: "Ada"

• _naay_version is validated but not enforced – it defaults to ysaqml.DEFAULT_NAAY_VERSION and is kept intact when re-saving.
• rows is an ordered list of dictionaries where every value is a string.

ysaqml serializes Python None values as the sentinel string __NULL__. When loading back, that string becomes None again, so avoid storing it as a literal value.

Testing

pytest

The tests create temporary directories and exercise the full load → mutate → dump loop.