datastore
simple, unified API for multiple data stores
datastore is a generic layer of abstraction for data store and database access. It is a simple API with the aim to enable application development in a datastore-agnostic way, allowing datastores to be swapped seamlessly without changing application code. Thus, one can leverage different datastores with different strengths without committing the application to one datastore throughout its lifetime. It looks like this:
+---------------+
| application | <--- No cumbersome SQL or Mongo specific queries!
+---------------+
| <--- simple datastore API calls
+---------------+
| datastore | <--- datastore implementation for underlying db
+---------------+
| <--- database specific calls
+---------------+
| various dbs | <--- MySQL, Redis, MongoDB, FS, ...
+---------------+
In addition, grouped datastores significantly simplify interesting data access patterns (such as caching and sharding).
About
Install
From pypi (using pip):
sudo pip install datastore
From pypi (using setuptools):
sudo easy_install datastore
From source:
git clone https://github.com/datastore/datastore/
cd datastore
sudo python setup.py install
Subprojects
- datastore.aws - aws s3 implementation
- datastore.git - git implementation
- datastore.mongo - monogdb implementation
- datastore.memcached - memcached implementation
- datastore.pylru - pylru cache implementation
- datastore.redis - redis implementation
- datastore.leveldb - leveldb implementation
Documentation
The documentation can be found at: http://datastore.readthedocs.org/en/latest/
License
datastore is under the MIT License.
Contact
datastore is written by Juan Batiz-Benet. It was originally part of py-dronestore. On December 2011, it was re-written as a standalone project.
Project Homepage: https://github.com/datastore/datastore
Feel free to contact me. But please file issues in github first. Cheers!
Contributing
Implementations
Please write and contribute implementations for other data stores. This project can only be complete with lots of help.
Style
Please follow proper pythonic style in your code.
See PEP 8 and the Google Python Style Guide.
Docs
Please document all code. datastore uses sphinx for documentation. Take
a look at the docs/ directory.
To make sure the documentation compiles, run:
cd docs make html open .build/html/index.html
Which should -- if all goes well -- open your favorite browser on the newly-built docs.
Examples
Hello World
>>> import datastore.core
>>> ds = datastore.DictDatastore()
>>>
>>> hello = datastore.Key('hello')
>>> ds.put(hello, 'world')
>>> ds.contains(hello)
True
>>> ds.get(hello)
'world'
>>> ds.delete(hello)
>>> ds.get(hello)
None
Hello filesystem
>>> import datastore.filesystem
>>>
>>> ds = datastore.filesystem.FileSystemDatastore('/tmp/.test_datastore')
>>>
>>> hello = datastore.Key('hello')
>>> ds.put(hello, 'world')
>>> ds.contains(hello)
True
>>> ds.get(hello)
'world'
>>> ds.delete(hello)
>>> ds.get(hello)
None
Hello Tiered Access
>>> import pymongo
>>> import datastore.core
>>>
>>> from datastore.mongo import MongoDatastore
>>> from datastore.pylru import LRUCacheDatastore
>>> from datastore.filesystem import FileSystemDatastore
>>>
>>> conn = pymongo.Connection()
>>> mongo = MongoDatastore(conn.test_db)
>>>
>>> cache = LRUCacheDatastore(1000)
>>> fs = FileSystemDatastore('/tmp/.test_db')
>>>
>>> ds = datastore.TieredDatastore([cache, mongo, fs])
>>>
>>> hello = datastore.Key('hello')
>>> ds.put(hello, 'world')
>>> ds.contains(hello)
True
>>> ds.get(hello)
'world'
>>> ds.delete(hello)
>>> ds.get(hello)
None
Hello Sharding
>>> import datastore.core
>>>
>>> shards = [datastore.DictDatastore() for i in range(0, 10)]
>>>
>>> ds = datastore.ShardedDatastore(shards)
>>>
>>> hello = datastore.Key('hello')
>>> ds.put(hello, 'world')
>>> ds.contains(hello)
True
>>> ds.get(hello)
'world'
>>> ds.delete(hello)
>>> ds.get(hello)
None
API
The datastore API places an emphasis on simplicity and elegance. Only four core methods must be implemented (get, put, delete, query).
get(key)
Return the object named by key or None if it does not exist.
Args:
key: Key naming the object to retrieve
Returns:
object or None
put(key, value)
Stores the object value named by key.
How to serialize and store objects is up to the underlying datastore.
It is recommended to use simple objects (strings, numbers, lists, dicts).
Args:
key: Key naming `value`
value: the object to store.
delete(key)
Removes the object named by key.
Args:
key: Key naming the object to remove.
query(query):
Returns an iterable of objects matching criteria expressed in query
Implementations of query will be the largest differentiating factor
amongst datastores. All datastores must implement query, even using
query's worst case scenario, see Query class for details.
Args:
query: Query object describing the objects to return.
Returns:
iterable cursor with all objects matching criteria
Specialized Features
Datastore implementors are free to implement specialized features, pertinent only to a subset of datastores, with the understanding that these should aim for generality and will most likely not be implemented across other datastores.
When implementings such features, please remember the goal of this project: simple, unified API for multiple data stores. When making heavy use of a particular library's specific functionality, perhaps one should not use datastore and should directly use that library.
Key
A Key represents the unique identifier of an object.
Our Key scheme is inspired by file systems and the Google App Engine key model.
Keys are meant to be unique across a system. Keys are hierarchical, incorporating increasingly specific namespaces. Thus keys can be deemed 'children' or 'ancestors' of other keys.
Key('/Comedy')
Key('/Comedy/MontyPython')
Also, every namespace can be parametrized to embed relevant object
information. For example, the Key name (most specific namespace) could
include the object type:
Key('/Comedy/MontyPython/Actor:JohnCleese')
Key('/Comedy/MontyPython/Sketch:CheeseShop')
Key('/Comedy/MontyPython/Sketch:CheeseShop/Character:Mousebender')
