@jsonapi

Decorate a view function that talks JSON.

Such function can return:

• tuples of (response, status[, headers]): to set custom status code and optionally - headers

• Instances of `JsonResponse <#jsonresponse>`__

• The result of helper function `make_json_response <#make_json_response>`__

Example:

from flask.ext.jsontools import jsonapi

@app.route('/users')
@jsonapi
def list_users():
    return [
        {'id': 1, 'login': 'kolypto'},
        #...
    ]

@app.route('/user/<int:id>', methods=['DELETE'])
def delete_user(id):
    return {'error': 'Access denied'}, 403

from flask.ext.jsontools import jsonapi

@app.route('/user', methods=['GET'])
@jsonapi
def list_users():
    return [ { 1: 'first', 2: 'second' } ]

@app.route('/user/<int:id>', methods=['GET'])
@jsonapi
def get_user(id):
    return list_users().get_json()[id]  # Long form
    return list_users()[id]  # Shortcut

DynamicJSONEncoder

In python, de-facto standard for encoding objects of custom classes is the __json__() method which returns the representation of the object.

DynamicJSONEncoder is the implementation of this protocol: if an object has the __json__() method, its result if used for the representation.

You’ll definitely want to subclass it to support other types, e.g. dates and times:

from flask.ext.jsontools import DynamicJSONEncoder

class ApiJSONEncoder(DynamicJSONEncoder):
    def default(self, o):
        # Custom formats
        if isinstance(o, datetime.datetime):
            return o.isoformat(' ')
        if isinstance(o, datetime.date):
            return o.isoformat()
        if isinstance(o, set):
            return list(o)

        # Fallback
        return super(DynamicJSONEncoder, self).default(o)

Now, just install the encoder to your Flask:

from flask import Flask

app = Flask(__name__)
app.json_encoder = DynamicJSONEncoder

JsonSerializableBase

Serializing SqlAlchemy models to JSON is a headache: if an attribute is present on an instance, this does not mean it’s loaded from the database.

JsonSerializableBase is a mixin for SqlAlchemy Declarative Base that adds a magic __json__() method, compatible with `DynamicJSONEncoder <#dynamicjsonencoder>`__. When serializing, it makes sure that entity serialization will never issue additional requests.

Example:

from sqlalchemy.ext.declarative import declarative_base
from flask.ext.jsontools import JsonSerializableBase

Base = declarative_base(cls=(JsonSerializableBase,))

class User(Base):
    #...

Now, you can safely respond with SqlAlchemy models in your JSON views, and jsontools will handle the rest :)

MethodView

Using MethodView class for methods, decorate them with @methodview(), which takes the following arguments:

• methods=(): Iterable of HTTP methods to use with this method.

• ifnset=None: Conditional matching. List of route parameter names that should not be set for this method to match.

• ifset=None: Conditional matching. List of route parameter names that should be set for this method to match.

This allows to map HTTP methods to class methods, and in addition define when individual methods should match.

Quick example:

from flask.ext.jsontools import jsonapi, MethodView, methodview

class UserView(MethodView):
    # Canonical way to specify decorators for class-based views
    decorators = (jsonapi, )

    @methodview
    def list(self):
        """ List users """
        return db.query(User).all()

    @methodview
    def get(self, user_id):
        """ Load a user by id """
        return db.query(User).get(user_id)

userview = CrudView.as_view('user')
app.add_url_rule('/user/', view_func=userview)
app.add_url_rule('/user/<int:user_id>', view_func=userview)

Now, GET HTTP method is routed to two different methods depending on conditions. Keep defining more methods to get good routing :)

To simplify the last step of creating the view, there’s a helper:

UserView.route_as_view(app, 'user', ('/user/', '/user/<int:user_id>'))

RestfulView

Since MethodView is mostly useful to expose APIs over collections of entities, there is a RESTful helper which automatically decorates some special methods with @methodview.

By subclassing RestfulView and implementing some of these methods, you’ll get a complete API endpoint with a single class.

It’s also required to define the list of primary key fields by defining the primary_key property:

from flask.ext.jsontools import jsonapi, RestfulView

class User(RestfulView):
    decorators = (jsonapi, )
    primary_key = ('id',)

    #region Operation on the collection

    def list():
        return db.query(User).all()

    def create():
        db.save(user)
        return user

    #endregion

    #region Operation on entities

    def get(id):
        return db.query(User).get(id)

    def replace(id):
        db.save(user, id)

    def update(id):
        db.save(user)

    def delete(id):
        db.delete(user)

    #endregion

When a class like this is defined, its metaclass goes through the methods and decorates them with @methodview. This way, list() gets @methodview('GET', ifnset=('id',)), and get() gets @methodview('GET', ifset=('id',)).