- Author
- Jefferson Heard
- Copyright
- 2015 Jefferson Heard
- License
- Apache 2.0
- Development home
- https://github.com/jeffersonheard/sondra
- Development status
- beta
- Maintainer
- Jefferson Heard
- Documentation
- https://jeffersonheard.github.io/sondra
Sondra is an "ORM" and REST-ful webservice framework for Python 3.x, Flask, and RethinkDB with some unique features. Sondra's goal is to aid full stack developers by letting them focus on data models and functionality instead of writing workarounds and glue code. It embraces common "shortcuts" developers take in common full-stack web applications, e.g. merging "Model" and "Controller" in the oft-used MVC pattern.
Sondra does not currently support asynchronous access to RethinkDB. The goal is to eventually support Tornado
- A clear, DRY heirarchical application structure that emphasizes convention over configuration.
- Authentication via JSON Web Tokens (JWT)
- JSON-Schema validation for documents.
- Expose methods on documents, collections, and applications, complete with schemas for call and return.
- A clear, predictable URL scheme for all manner of API calls, covering a broad set of use-cases.
- Self documenting APIs with both human-readable help based on docstrings and schemas for every call.
- Use API idiomatically over HTTP and native Python without writing boilerplate code
$ python setup.py
Contributions are welcome. I haven't gotten around to posting a "master plan" of any sort yet, but I will get there. If you're interested in contributing, please email me directly or simply fork the GitHub project and issue a pull request.
For support, please see the project page on GitHub
This project is licensed under the Apache license V2. See LICENSE.
A Sondra API is exposed in Flask as a suite of applications. Each application contains a number of document collections, and each collection contains a number of documents.
Sondra tries to take advantage of all manner of Python idioms in a sane manner. It generates as much as possible, while avoiding "magic tricks" or introducing conventions that are already covered by Python idioms. This means that:
- Online documentation is generated at every level from reStructuredText or Google style docstrings.
- Method schemas are generated from annotated function signatures
- All URL features that urllib.parse recognizes are taken advantage of to create a regular URL scheme that encompasses all manner of calls.
Sondra uses a regular URL scheme for its web services. See the docs directory for more details:
- Underscores in Python identifiers are replaced with more web-friendly dashes (slugs)
- Paths address app / collection / instance (document)
- The last path element include an optional . (dot), which separates the object being addressed and a method being called.
- Fragments allow the caller to address sub-documents in JSON responses. Fragments delimted with # are not passed by browsers to the server. The alternative for delimiting fragments is @!
- Path parameters (on the last element, with a ";", see the URL spec or urlparse docs) are used to specify output
format or content directives. Currently supported:
- json or format=json retrieves data in JSON format.
- geojson or format=geojson retrieves data in GeoJSON feature or feature collection format.
- help or format=help retrieves HTML autogenerated help.
- schema or format=schema retrieves a JSON-Schema document for the service call.
A Document is a single document that conforms to a JSON-Schema object type. That is, it is never a simple type nor an array of items.
Documents may expose methods to the HTTP api. These are similar to instance methods in Python. They operate on an individual document in a collection instead. Document methods might include operations that combine multiple documents to make a third (add, multiply, divide, subtract, or similar) or they might provide specific views of a document. Anything that you would write as an "instance method" in Python.
A Collection is a RethinkDB table that contains a specific subclass of Document, which is defined by a single JSON-Schema. The collection class defines additionally:
- The primary key name (defaults to the RethinkDB default of "id")
- Indexes
- Any document properties that require "special treatment" in RethinkDB such as geographical and date/time types.
- Relations to other Collections
- The Application class it belongs to.
Collections may expose methods to the HTTP api. These are similar to class methods in Python, as they operate on the collection itself and not the individual documents. Collection methods might provide special filtering, create documents according to a specific template, or set properties on the collection itself. Anything you would write as a "class method" in Python
An Application is a reusable grouping of collections and a set of optional application methods, which operate a bit like globally available functions. Applications are bound to a single database within RethinkDB.
Applications may expose methods to the HTTP api. These are similar to the functions that are defined at the module level in Python. They are not specific to a particular class or instance, but instead are defined to provide broad functionality for the whole application.
A Suite defines the environment of applications, including database connections and provides some basic functionality. Every application is registered with the global Suite object, which itself implements Python's Mapping protocol to provide dictionary-like lookup of application objects. The "Suite" object determines the base path of all Application APIs. Suites are similar in nature to Django's settings.py except that they are class-based. There may be only one concrete class of Suite in your Flask app, although it may derive from any number of abstract Suite mixins.