Ormophine

The most Pythonic ORM. Fast, intuitive, and gets out of your way.

Write database queries the way you think — in plain Python.

Documentation & AI Assistance

📖 Full Documentation
Comprehensive guides, API references, and examples are available at:
👉 https://ormophine.readthedocs.io/en/latest/index.html

🤖 AI-Powered Assistance
To help you write queries and debug your code, Ormophine ships with AI reference files (Sqlite.AI.Reference.txt, MySQL.AI.Reference.txt, PostgreSQL.AI.Reference.txt).

You can find these files in the root directory of the installed package. Simply attach the appropriate file to ChatGPT, Claude, or Gemini, ask your question, and the AI will respond using the exact API and behavior of your Ormophine version. It's like having an Ormophine expert on standby!

The Problem with Other ORMs

Most Python ORMs are either too verbose, too magical, or require too much boilerplate. Compare the everyday workflow of connecting, accessing tables, inserting data, and querying using popular ORMs versus Ormophine.

Connecting to the Database

SQLAlchemy:

from sqlalchemy import create_engine
engine = create_engine('sqlite:///my_db.db')

PonyORM:

from pony.orm import Database
db = Database()
db.bind(provider='sqlite', filename='my_db.db', create_db=True)

Peewee:

from peewee import SqliteDatabase
db = SqliteDatabase('my_db.db')

Ormophine:

from Ormophine.Sqlite import Driver

db = Driver('my_db.db')

Accessing Tables

SQLAlchemy: Requires manual reflection or pre-defined models

from sqlalchemy import Table, MetaData
metadata = MetaData()
users = Table('users', metadata, autoload_with=engine)

PonyORM: Requires defining entities and generating mappings

from pony.orm import Required
class User(db.Entity):
    name = Required(str)
    age = Required(int)
db.generate_mapping(create_tables=True)

Peewee: Requires defining models and explicitly linking them to the database

from peewee import Model, CharField, IntegerField
class User(Model):
    name = CharField()
    age = IntegerField()
    class Meta:
        database = db

Ormophine:

# Tables and columns are discovered and mapped dynamically as attributes
users = db.users

Inserting Data

SQLAlchemy: Requires explicit connection context and commit

with engine.connect() as conn:
    conn.execute(users.insert().values(
        name='Alice', 
        email='alice@example.com', 
        age=30
    ))
    conn.commit()

PonyORM: Requires explicit db_session context

from pony.orm import db_session
with db_session:
    User(name='Alice', email='alice@example.com', age=30)

Peewee: Requires calling .execute() on the query construct

User.insert(
    name='Alice', 
    email='alice@example.com', 
    age=30
).execute()

Ormophine:

# Auto-committed, uses Pythonic dictionary mapping with actual column objects
users.insert({
    users.name:  'Alice',
    users.email: 'alice@example.com',
    users.age:   30
})

Fetching Data with Complex Conditions

Let's try to fetch rows where the lowercased name starts with 'ab', AND a specific slice of the lastname equals 'connor', ordered by age.

SQLAlchemy: Verbose function calls and manual string manipulation for slicing

from sqlalchemy import select, func
stmt = select(users.c.name, users.c.age).where(
    func.lower(users.c.name).like('ab%'),
    func.substr(users.c.lastname, 6, func.length(users.c.lastname) - 7) == 'connor'
).order_by(users.c.age)
with engine.connect() as conn:
    results = conn.execute(stmt).fetchall()

PonyORM: Requires lambda functions and lacks intuitive slicing

from pony.orm import db_session, select
with db_session:
    query = select(u for u in User if u.name.lower().startswith('ab'))
    # String slicing like [5:-2] is not natively supported in PonyORM queries
    query = query.order_by(lambda u: u.age)
    results = [(u.name, u.age) for u in query]

Peewee: Uses SQL function wrappers and lacks native Python slicing

from peewee import fn
# Peewee lacks native string slicing in ORM queries
query = User.select(User.name, User.age).where(
    fn.LOWER(User.name).startswith('ab')
    # User.lastname[5:-2] == 'connor' is not possible natively
).order_by(User.age)
results = list(query.dicts())

Ormophine:

# Pure Python syntax! Slicing and string methods translate directly to SQL under the hood.
rows = users.get_row(
    [users.name, users.age],
    where=(users.name.lower().startswith('ab')) & (users.lastname[5:-2] == 'connor'),
    order_by=users.age
)

Atomic / Batch Transactions

Performing multiple write operations in a single, atomic transaction is crucial for data integrity and speed. Let's insert 2 users, update 1, and delete 1.

SQLAlchemy: Requires explicit connection block and manual execution for each statement

with engine.begin() as conn:
    conn.execute(users.insert().values(name='Dave', email='dave@example.com', age=40))
    conn.execute(users.insert().values(name='Eve', email='eve@example.com', age=28))
    conn.execute(users.update().where(users.c.name == 'Alice').values(age=31))
    conn.execute(users.delete().where(users.c.name == 'Bob'))

PonyORM: Requires db_session context and imperative object manipulation for updates/deletes

from pony.orm import db_session
with db_session:
    User(name='Dave', email='dave@example.com', age=40)
    User(name='Eve', email='eve@example.com', age=28)
    alice = User.get(name='Alice')
    if alice: alice.age = 31
    bob = User.get(name='Bob')
    if bob: bob.delete()

Peewee: Requires atomic context and explicit .execute() on every query construct

with db.atomic():
    User.insert(name='Dave', email='dave@example.com', age=40).execute()
    User.insert(name='Eve', email='eve@example.com', age=28).execute()
    User.update(age=31).where(User.name == 'Alice').execute()
    User.delete().where(User.name == 'Bob').execute()

Ormophine:

# Fluent batch builder: queues operations and commits atomically
batch = users.batch()
batch.insert({users.name: 'Dave', users.email: 'dave@example.com', users.age: 40})
batch.insert({users.name: 'Eve', users.email: 'eve@example.com', users.age: 28})
batch.update({users.age: 31}, where=users.name == 'Alice')
batch.delete_row(where=users.name == 'Bob')
batch.run() # Executes all and commits in one transaction

Same results. No boilerplate. No complex function mapping. Just Python.

Why Ormophine?

• Intuitive Pythonic Syntax — columns behave like native Python variables. We designed Ormophine to simulate standard Python string and sequence behaviors directly in SQL. Instead of learning a new DSL or using verbose SQL functions, you just write Python, and Ormophine translates it into optimized, parameterized SQL under the hood:
  • String Concatenation: Use the standard Python + operator to concatenate string columns and literals seamlessly.
  • Native String Methods: Chain Python string methods like lower(), upper(), strip(), lstrip(), rstrip(), replace(), startswith(), and endswith() directly on column objects.
  • Sequence Slicing: Use Python's native slice syntax (e.g., column[2:5] or column[-4:]) to extract substrings, which automatically translates to native SQL substring functions.
  • Logical & Arithmetic Operators: Combine conditions using Python's bitwise operators (&, |) and perform arithmetic (+, -, *, /) just like regular Python variables.
• Fast & Thread-Safe — built on a dedicated writer queue (SQLite) and robust connection pooling (MySQL/PostgreSQL); parallel reads, serialized writes.
• Fault-Tolerant Connection Pools — automatically detects broken connections (e.g., database restarts) and seamlessly recreates them without crashing your application.
• Multi-database — one unified API across SQLite, MySQL, and PostgreSQL. Switch databases by changing your import.
• Dynamic Schema Mapping — tables and columns are discovered automatically and attached to the driver instance.
• Built-in DB Administration — manage users, permissions, create/drop databases, and run maintenance tasks (like PostgreSQL VACUUM or SQLite PRAGMA) directly from the driver.
• WAL mode support (SQLite) — automatic checkpointing for maximum write throughput.

⚡ Benchmark Results

To demonstrate Ormophine's raw performance, we benchmarked it against popular Python ORMs (SQLAlchemy, PonyORM, and Peewee) across SQLite, PostgreSQL, and MySQL.

Methodology

We evaluate two distinct scenarios to measure both transactional overhead and bulk efficiency:

1. Single Operations: Measures the time taken to execute CRUD queries where a COMMIT is issued immediately after every single insert, update, and delete. This tests the ORM's baseline overhead and connection management for isolated transactions.
2. Batch Operations: Measures the time taken to execute a block of CUD (Create, Update, Delete) queries where all statements are executed first, and a single COMMIT is issued at the end. This tests the ORM's efficiency in bulk transactional processing.

Note on Variance & Equivalence: Due to natural system fluctuations, each test run can have a variance of up to ±10%. Therefore, performance differences of less than 5% are considered statistically insignificant (margin of error). In the charts below, differences under 5% are displayed in gray and marked as "≈ Equal", rather than claiming a marginal advantage.

You can access the benchmark Jupyter notebooks in the project repository at Ormophine/{Sqlite, Postgresql, Mysql}/Benchmark to run the tests on your own hardware.

You can also use this Google Colab notebooks:

Sqlite: https://colab.research.google.com/drive/1KK3sr8H_Crd29fmnq3VmpmE88aLNT3Yr?usp=sharing

MySQL: https://colab.research.google.com/drive/1ndwmN0C9UTZHTNmLh8-fT9rEg-DSrzHQ?usp=sharing

PostgeSQL: https://colab.research.google.com/drive/1XYrC30vUciS1YgY6M5MBoxwO9YTltzkD?usp=sharing

PostgreSQL Results

Single Operations Test (Executed 10,000 queries total — 200 repeats × 50 chunk size — for each CRUD operation per ORM)

Inserts:

Updates:

Reads:

Deletes:

Batch Operation Test (Executed 500 queries total — 5 repeats × 100 statements per chunk — for each CUD operation per ORM)

Inserts:

Updates:

Deletes:

Sqlite Results

Single Operations Test (Executed 10,000 queries total — 200 repeats × 50 chunk size — for each CRUD operation per ORM)

Inserts:

Updates:

Reads:

Deletes:

Batch Operation Test (Executed 500 queries total — 5 repeats × 100 statements per chunk — for each CUD operation per ORM)

Inserts:

Updates:

Deletes:

MySQL Results

Single Operations Test (Executed 10,000 queries total — 200 repeats × 50 chunk size — for each CRUD operation per ORM)

Inserts:

Updates:

Reads:

Deletes:

Batch Operation Test (Executed 500 queries total — 5 repeats × 100 statements per chunk — for each CUD operation per ORM)

Inserts:

Updates:

Deletes:

Quick Examples

Connect and access tables

from Ormophine.Sqlite import Driver

db = Driver('company.db')

# Tables and columns are discovered automatically
users   = db.users
orders  = db.orders

Insert

users.insert({
    users.name:  'Alice',
    users.email: 'alice@example.com',
    users.age:   30
})

Select with conditions

name, email, age = users.name, users.email, users.age

rows = users.get_row(
    [name, email],
    where   = (age >= 18) & name.startswith('A'),
    order_by = age
)

Update

users.update(
    update = {users.age: users.age + 1},
    where  = users.status == 'active'
)

Bulk insert

users.bulk_insert(
    columns   = [users.name, users.age],
    data_list = [['Bob', 25], ['Carol', 32], ['Dave', 28]]
)

Joins

from Ormophine.Sqlite import Join

result = orders.join(
    columns    = [users.name, orders.amount, orders.date],
    joins_list = [Join.Inner(users, users.id == orders.user_id)],
    where      = orders.amount > 100,
    order_by   = [orders.date]
)

Schema management

from Ormophine.Sqlite import TableStructure, DataTypes

schema = TableStructure('products', strict=True)
schema.add_column('id',    DataTypes.INTEGER(), primary_key=True)
schema.add_column('title', DataTypes.TEXT(max_length=100), not_null=True, unique=True)
schema.add_column('price', DataTypes.REAL(), default_value=0.0)

products = db.create_table(schema)

# Add / rename / drop columns dynamically
products.add_column('stock', DataTypes.INTEGER(), default_value=0, not_null=True)
products.rename_column(products.stock, 'inventory')
products.delete_column(products.inventory, True, True, True)

Safe Deletion & Administration

# Triple-confirmation flags prevent catastrophic accidental drops
db.delete_table(db.users, are_you_sure=True, are_you_really_sure=True, for_sure=True)

# Create a new database on the fly during connection (MySQL/PostgreSQL)
# from Ormophine.Mysql import Driver
# db = Driver(host='localhost', port=3306, username='root', password='pass', db_name='new_db', create_new_db=True)

WAL mode and performance tuning (SQLite)

db.set_WAL_mode(True, wal_timer=60)   # automatic checkpoint every 60 s

db.SetPragma.synchronous('NORMAL')
db.SetPragma.cache_size(-4000)        # 4 MiB page cache
db.SetPragma.foreign_keys(True)

Operator Reference

Ormophine columns support native Python expressions — all values are automatically parameterized to prevent SQL injection.

Installation

pip install Ormophine

Project Status & Roadmap

⚠️ Work in Progress

Ormophine is currently in active development. While it is highly functional and fast, it is not yet as feature-complete or massive in size as legacy ORMs like SQLAlchemy or Django ORM.

Our philosophy is to keep the core lightweight and fast. In future releases, we plan to simulate even more Python string and list methods to make the query syntax even closer to pure Python.

Current Roadmap

• SQLite backend with full ORM
• MySQL backend with connection pooling
• PostgreSQL backend with connection pooling
• Operator overloading and slicing ([]) for columns
• String methods simulation (lower, upper, strip, startswith, etc.)
• Read-only connection pool / Non-blocking reads
• WAL mode + automatic checkpointing
• Batch / bulk operations
• AI Reference files for LLM assistance
• Expanding simulated Python methods (.replace(), .find(), etc.)
• Video Tutorials
• Benchmark suite publication
• Async support

Video Tutorials

🎥 Coming Soon! We are preparing a comprehensive video series to help you get started with Ormophine, from basic connections to advanced concurrent read/write pooling and schema management.

Stay tuned—links will be posted here soon.

Contributing

The codebase is currently in active development. Contributions, bug reports, and feature requests are very welcome! Please feel free to open an issue or submit a pull request.

License

MIT — free to use, modify, and distribute.