Skip to content

mplanchard/falcon-marshmallow

Repository files navigation

falcon-marshmallow

image

image


Marshmallow serialization/deserialization middleware for Falcon

Maintained By Matthew Planchard
Author Matthew Planchard
License MIT
Compatibility Python 2.7+, 3.4+
Contributors @timc13, Your Name Here!

Note: this package is a fork of ihiji/falcon-marshmallow, of which I am the original author. The project was abandoned after Ihiji was purchased, so I'm maintaining it here now. Issues opened on the original fork will not be seen or resolved.

Installation

Install from pypi with:

pip install falcon_marshmallow

Usage

The primary middleware provided by this package is called Marshmallow. To use it, simply add it to your Falcon app instantiation:

from falcon import API
from falcon_marshmallow import Marshmallow

app = API(
    middleware=[
        Marshmallow(),
    ]
)

The Marshmallow middleware looks for schemas defined on your resources, and, if it finds an applicable schema, uses it to serialize incoming requests and deserialize outgoing responses. Schema attributes should either be named schema or <method>_schema, where <method> is an HTTP method. If both an appropriate method schema and a general schema are defined, the method schema takes precedence.

Marshmallow assumes JSON serialization and uses simplejson as the default (de)serializer, but if you specify a different serialization module in a schema's Meta class, that will be seamlessly integrated into this library's (de)serialization.

By default, if no schema is found, the Marshmallow middleware will still attempt to (de)serialize data using the simplejson module. This can be disabled when instantiating the middleware by setting force_json to False.

Two extra middleware classes are provided for convenience:

  • JSONEnforcer raises an HTTPNotAcceptable error if the client request indicates that it will does not accept JSON and ensures that the Content-Type of requests is "application/json" for specified HTTP methods (default PUT, POST, PATCH).
  • EmptyRequestDropper returns an HTTPBadRequest if a request has a non-zero Content-Length header with an empty body

Examples

Standard ReST Resource

Let's look at a standard ReST resource corresponding to a Philosopher resource in our chosen data store:

from datetime import date
from random import randint

from marshmallow import fields, Schema
from falcon import API
from falcon_marshmallow import Marshmallow
from wsgiref import simple_server


class MyDataStore:
    """Whatever DB driver you like"""

    def get(self, table, phil_id):
        print('I got item with id %s from %s' % (phil_id, table))
        return {
            'id': phil_id,
            'name': 'Albert Camus',
            'birth': date(1913, 11, 7),
            'death': date(1960, 1, 4),
            'schools': ['existentialism', 'absurdism'],
            'works': ['The Stranger', 'The Myth of Sissyphus']
        }

    def insert(self, table, obj):
        print('I inserted %s into %s' % (obj, table))
        return {
            'id': randint(1, 100),
            'name': 'Søren Kierkegaard',
            'birth': date(1813, 5, 5),
            'death': date(1855, 11, 11),
            'schools': ['existentialism'],
            'works': ['Fear and Trembling', 'Either/Or']
        }


class Philosopher(Schema):
    """Philosopher schema"""
    id = fields.Integer()
    name = fields.String()
    birth = fields.Date()
    death = fields.Date()
    schools = fields.List(fields.String())
    works = fields.List(fields.String())


class PhilosopherResource:

    schema = Philosopher()

    def on_get(self, req, resp, phil_id):
        """req['result'] will be automatically serialized

        The key in which results are stored can be customized when
        the middleware is instantiated.
        """
        req.context['result'] = MyDataStore().get('philosophers', phil_id)


class PhilosopherCollection:

    schema = Philosopher()

    def on_post(self, req, resp):
        """req['json'] contains our deserialized data

        The key in which deserialized data can be stored can be
        customized when the middleware is instantiated.
        """
        inserted = MyDataStore().insert('philosophers', req.context['json'])
        req.context['result'] = inserted


app = API(middleware=[Marshmallow()])

app.add_route('/v1/philosophers', PhilosopherCollection())
app.add_route('/v1/philosophers/{phil_id}', PhilosopherResource())


if __name__ == '__main__':
    svr = simple_server.make_server('127.0.0.1', 8080, app)
    svr.serve_forever()

Done!

When parsing a request body, if it cannot be decoded or its JSON is malformed, an HTTPBadRequest error will be raised. If the deserialization of the request body fails due to schema validation errors, an HTTPUnprocessableEntity error will be raised.

We can test our new server easily enough using the requests library:

>>> import requests

# - GET some philosopher - #

>>> resp = requests.get('http://127.0.0.1:8080/v1/philosophers/12')

>>> resp.text
'{"birth": "1913-11-07", "id": 12, "death": "1960-01-04", "works": ["The Stranger", "The Myth of Sissyphus"], "schools": ["existentialism", "absurdism"], "name": "Albert Camus"}'

>>> resp.json()
{'birth': '1913-11-07',
 'death': '1960-01-04',
 'id': 12,
 'name': 'Albert Camus',
 'schools': ['existentialism', 'absurdism'],
 'works': ['The Stranger', 'The Myth of Sissyphus']}

# - POST a new philosopher - #

>>> post_data = resp.json()

>>> import json

>>> presp = requests.post('http://127.0.0.1:8080/v1/philosophers', data=json.dumps(post_data))

>>> presp.json()
{'birth': '1813-05-05',
 'death': '1855-11-11',
 'id': 100,
 'name': 'Søren Kierkegaard',
 'schools': ['existentialism'],
 'works': ['Fear and Trembling', 'Either/Or']}


# - Try to POST bad data - #

>>> post_data['birth'] = 'not a date'

>>> presp = requests.post('http://127.0.0.1:8080/v1/philosophers', data=json.dumps(post_data))

>>> presp
<Response [422]>

>>> presp.json()
{'description': '{"birth": ["Not a valid date."]}',
 'title': '422 Unprocessable Entity'}

Customization

Customization is effected by keyword arguments to the middleware constructor. The constructor takes the following arguments:

  • req_key (default json) - the key on the request's context dict on which to store parsed request data
  • resp_key (default result) - the key on the request's context dict in which data to be serialized for a response should be stored
  • force_json (default True) - attempt to (de)serialize request and response bodies to/from JSON even if no schema is defined for a resource
  • json_module (default simplejson) - the module to use for (de)serialization; must implement the public interface of the json standard library module

A Note on Python 2

Python 2 will be at its End of Life (EoL) at the end of 2019. This package is relatively simple, though, and Python 2 compatibility is therefore not a huge burden to maintain. As such, we will continue to maintain Python 2 compatibility after its EoL, until and unless we decide that maintaining Python 2 compatibility puts our users at risk due to potential security vulnerabilities, we feel that the burden of maintaining Python 2 compatibility has become too high for our maintainers, or we determine that we can provide a significant increase in value by dropping Python 2support.

In any of those cases, we will try to clearly communicate the change with a major version bump.

Contributing

Contributions are welcome. Please feel free to raise Issues, submit PRs, fix documentation typos, etc. If opening a PR, please be sure to run tests, and ensure that your additions are compatible with Python 2.7, 3.4, and above.

Ideally, PRs should have tests, but feel free to open a PR with or without them. The maintainers will either suggest some tests for you to add, or, if you are not able to add tests yourself, we may open a PR against your branch with some added tests before merging.

Development requires that you have Python 3 available on your path.

Development

To set up a local virtual environment with all required packages installed, run:

make setup

If you are using VSCode, the .vscode/settings.json file included in this project should now be automatically configured to autoformat on save and to perform all of the lint checks that are required for this package.

Linting

The linting checks that run in CI can be manually run locally with:

make lint

Note that this will automatically create a local virtual environment for you if make setup has not yet been run.

Testing

To run tests against Python 2.7 and 3.4 forward, you can just run:

make test

Note that this will automatically create a local virtual environment for you if make setup has not yet been run.

Testing against all environments of course requires that you have the requisite Python executables available on your PATH. If you don't, you will get "interpreter not found" errors for the missing python versions.

To run against a particular version of Python, use, for example:

TESTENV=py37 make test-env

Where TESTENV is any of the environments configured in tox.ini, or any of tox's standard environments (e.g. py36, py37, etc.).