goatfish is a small, schemaless ORM that is backed by SQLite.

It's also this:

Its usage is very simple, just have your classes inherit from goatfish.Model and and specify a connection, and the goatfish methods are available to you. goatfish also supports querying for arbitrary properties in your models, as well as indexing on arbitrary properties. It does not enforce a schema of any kind.

It appears that this method is identical to what FriendFeed used to implement a schemaless layer over MySQL, which is pretty significant validation:

http://backchannel.org/blog/friendfeed-schemaless-mysql

To use goatfish, all you need to do is create a class that inherits from goatfish.Model:

import goatfish
import sqlite3

db_connection = sqlite3.connect(":memory:")

class Test(goatfish.Model):
    class Meta:
        # This is so we know where to connect.
        connection = db_connection
        indexes = (
            ("foo",), # Index on just `foo`.
            ("foo", "bar"), # Index on both `foo` and `bar`, together.
        )

# Create the necessary tables. If they exist, do nothing.
Test.initialize()

foo = Test()
foo.foo = "hi"
foo.bar = "hello"
foo.save()

# Retrieve all elements.
>>> [test.bar for test in Test.all()]
['hello']

# Count the number of elements.
>>> Test.count(foo="hi")
1

# Run a query with parameters (slow, loads every item from the DB to check it).
>>> [test.bar for test in Test.find(bar="hello")]
['hello']

# This uses an index, so it's fast.
>>> [test.foo for test in Test.find(foo="hi"})]
['hi']

# Fast, because there's an index for when both fields are used together.
>>> [test.bar for test in Test.find(foo="hi", bar="hello")]
['hello']

# Run a query with a parameter that doesn't exist in the dataset.
>>> [test.bar for test in Test.find({bar="hello", baz="hi"})]
[]

>>> Test.find_one(bar="hello").foo
"hi"

>>> print(Test.find_one(bar="doesn't exist"))
None

# Delete the element.
>>> foo.delete()

# Try to retrieve all elements again.
>>> [test.bar for test in Test.find()]
[]

What sets goatfish apart from other modules such as shelve, zodb, etc is its ability to query random attributes, and make those queries faster by using SQLite indexes.

The way this is achieved is by creating an intermediate table for each index we specify. The index tables consist of the uuid column, and one column for every field in the index. This way, we can store the value itself in these index tables and query them quickly, as the rows have SQLite indexes themselves.

The find() method uses these indexes automatically, if they exist, to avoid sequential scans. It will automatically use the largest index that contains the data we want to query on, so a query of {"foo": 3, "bar": 2} when only foo is indexed will use the index on foo to return the data, and do a sequential scan to match bar.

Right now, new indexes are only populated with data on save(), so you might miss rows when querying on indexes that are not ready yet. To populate indexes, go through the objects in your model and perform a save() in each of them. Convenience functions to populate single indexes will be provided shortly.

To install goatfish you need:

• Python 3.4 or later.

You have multiple options for installation:

• With pip (preferred), do pip install goatfish.
• With setuptools, do easy_install goatfish.
• To install from source, download it from https://github.com/stochastic-technologies/goatfish/ and do python setup.py install.

goatfish is distributed under the BSD license.