sqlstratum 0.1.0

Lightweight, source-first SQL AST + compiler + runner.

SQLStratum

SQLStratum is a modern, typed, deterministic SQL query builder and compiler for Python with a SQLite runner and a hydration pipeline. It exists to give applications and ORMs a reliable foundation layer with composable SQL, predictable parameter binding, and explicit execution boundaries.

Key Features

• Deterministic compilation: identical AST inputs produce identical SQL + params
• Typed, composable DSL for SELECT/INSERT/UPDATE/DELETE
• Safe parameter binding (no raw interpolation)
• Hydration targets for structured results
• SQLite-first execution via a small Runner API
• Testable compiled output and runtime behavior

Non-Goals

• Not an ORM (no identity map, relationships, lazy loading)
• Not a migrations/DDL system
• Not a full database abstraction layer for every backend yet (SQLite first)
• Not a SQL string templating engine

SQLStratum focuses on queries. DDL statements such as CREATE TABLE or ALTER TABLE are intended to live in a complementary library with similar design goals that is currently in the works.

Quickstart

import sqlite3

from sqlstratum import SELECT, INSERT, Table, col, Runner

users = Table(
    "users",
    col("id", int),
    col("email", str),
    col("active", int),
)

conn = sqlite3.connect(":memory:")
runner = Runner(conn)
runner.exec_ddl("CREATE TABLE users (id INTEGER PRIMARY KEY, email TEXT, active INTEGER)")

runner.execute(INSERT(users).VALUES(email="[e-mail]", active=1))
runner.execute(INSERT(users).VALUES(email="[e-mail]", active=0))

q = (
    SELECT(users.c.id, users.c.email)
    .FROM(users)
    .WHERE(users.c.active.is_true())
    .HYDRATE(dict)
)

rows = runner.fetch_all(q)
print(rows)

Why Table objects?

SQLStratum’s Table objects are the schema anchor for the typed, deterministic query builder. They provides column metadata and a stable namespace for column access, which enables predictable SQL generation and safe parameter binding. They also support explicit aliasing to avoid ambiguous column names in joins.

Project Structure

• AST: immutable query nodes in sqlstratum/ast.py
• Compiler: SQL + params generation in sqlstratum/compile.py
• Runner: SQLite execution and transactions in sqlstratum/runner.py
• Hydration: projection rules and targets in sqlstratum/hydrate.py

SQL Debugging

SQLStratum can log executed SQL statements (compiled SQL + parameters + duration), but logging is intentionally gated to avoid noisy output in production. Debug output requires two conditions:

• Environment variable gate: SQLSTRATUM_DEBUG must be truthy ("1", "true", "yes", case-insensitive).
• Logger gate: the sqlstratum logger must be DEBUG-enabled.

Why it does not work by default: Python logging defaults to WARNING level, so even if SQLSTRATUM_DEBUG=1 is set, DEBUG logs will not appear unless logging is configured.

To enable debugging in a development app:

Step 1 - set the environment variable:

SQLSTRATUM_DEBUG=1

Step 2 - configure logging early in the app:

import logging

logging.basicConfig(level=logging.DEBUG)
# or
logging.getLogger("sqlstratum").setLevel(logging.DEBUG)

Output looks like:

SQL: <compiled sql> | params={<sorted params>} | duration_ms=<...>

Architectural intent: logging happens at the Runner boundary (after execution). AST building and compilation remain deterministic and side-effect free, preserving separation of concerns.

Logo Inspiration

Vinicunca (Rainbow Mountain) in Peru’s Cusco Region — a high-altitude day hike from Cusco at roughly 5,036 m (16,500 ft). See Vinicunca for background.

Versioning / Roadmap

Current version: 0.1.0. Design notes and current limitations are tracked in NOTES.md. Roadmap planning is intentionally minimal at this stage and will evolve with real usage.

Authorship

Antonio Ognio is the maintainer and author of SQLStratum. ChatGPT is used for brainstorming, architectural thinking, documentation drafting, and project management advisory. Codex (CLI/agentic coding) is used to implement many code changes under Antonio's direction and review. The maintainer reviews and curates changes; AI tools are assistants, not owners, and accountability remains with the maintainer.

License

MIT License.

Contributing

PRs are welcome. Please read CONTRIBUTING.md for the workflow and expectations.