Read Latest Documentation - Browse GitHub Code Repository
hug aims to make developing Python driven APIs as simple as possible, but no simpler. As a result, it drastically simplifies Python API development.
hug's Design Objectives:
- Make developing a Python driven API as succinct as a written definition.
- The framework should encourage code that self-documents.
- It should be fast. A developer should never feel the need to look somewhere else for performance reasons.
- Writing tests for APIs written on-top of hug should be easy and intuitive.
- Magic done once, in an API framework, is better than pushing the problem set to the user of the API framework.
- Be the basis for next generation Python APIs, embracing the latest technology.
As a result of these goals, hug is Python 3+ only and built upon Falcon's high performance HTTP library
Get professionally supported hug with the Tidelift Subscription
Professional support for hug is available as part of the Tidelift Subscription. Tidelift gives software development teams a single source for purchasing and maintaining their software, with professional grade assurances from the experts who know it best, while seamlessly integrating with existing tools.
Installing hug is as simple as:
pip3 install hug --upgradeIdeally, within a virtual environment.
Build an example API with a simple endpoint in just a few lines.
# filename: happy_birthday.py
"""A basic (single function) API written using hug"""
import hug
@hug.get('/happy_birthday')
def happy_birthday(name, age:hug.types.number=1):
"""Says happy birthday to a user"""
return "Happy {age} Birthday {name}!".format(**locals())To run, from the command line type:
hug -f happy_birthday.pyYou can access the example in your browser at:
localhost:8000/happy_birthday?name=hug&age=1. Then check out the
documentation for your API at localhost:8000/documentation
Parameters can also be encoded in the URL (check
out happy_birthday.py for the whole
example).
@hug.get('/greet/{event}')
def greet(event: str):
"""Greets appropriately (from http://blog.ketchum.com/how-to-write-10-common-holiday-greetings/) """
greetings = "Happy"
if event == "Christmas":
greetings = "Merry"
if event == "Kwanzaa":
greetings = "Joyous"
if event == "wishes":
greetings = "Warm"
return "{greetings} {event}!".format(**locals())Which, once you are running the server as above, you can use this way:
curl http://localhost:8000/greet/wishes
"Warm wishes!"
# filename: versioning_example.py
"""A simple example of a hug API call with versioning"""
import hug
@hug.get('/echo', versions=1)
def echo(text):
return text
@hug.get('/echo', versions=range(2, 5))
def echo(text):
return "Echo: {text}".format(**locals())To run the example:
hug -f versioning_example.pyThen you can access the example from localhost:8000/v1/echo?text=Hi / localhost:8000/v2/echo?text=Hi Or access the documentation for your API from localhost:8000
Note: versioning in hug automatically supports both the version header as well as direct URL based specification.
hug's http method decorators don't modify your original functions. This makes testing hug APIs as simple as testing any other Python functions. Additionally, this means interacting with your API functions in other Python code is as straight forward as calling Python only API functions. hug makes it easy to test the full Python stack of your API by using the hug.test module:
import hug
import happy_birthday
hug.test.get(happy_birthday, 'happy_birthday', {'name': 'Timothy', 'age': 25}) # Returns a Response objectYou can use this Response object for test assertions (check
out test_happy_birthday.py ):
def tests_happy_birthday():
response = hug.test.get(happy_birthday, 'happy_birthday', {'name': 'Timothy', 'age': 25})
assert response.status == HTTP_200
assert response.data is not Nonehug exposes a __hug_wsgi__ magic method on every API module automatically. Running your hug based API on any standard wsgi server should be as simple as pointing it to module_name: __hug_wsgi__.
For Example:
uwsgi --http 0.0.0.0:8000 --wsgi-file examples/hello_world.py --callable __hug_wsgi__To run the hello world hug example API.
When building an API using the hug framework you'll use the following concepts:
METHOD Decorators get, post, update, etc HTTP method decorators that expose your Python function as an API while keeping your Python method unchanged
@hug.get() # <- Is the hug METHOD decorator
def hello_world():
return "Hello"hug uses the structure of the function you decorate to automatically generate documentation for users of your API. hug always passes a request, response, and api_version variable to your function if they are defined params in your function definition.
Type Annotations functions that optionally are attached to your methods arguments to specify how the argument is validated and converted into a Python type
@hug.get()
def math(number_1:int, number_2:int): #The :int after both arguments is the Type Annotation
return number_1 + number_2Type annotations also feed into hug's automatic documentation
generation to let users of your API know what data to supply.
Directives functions that get executed with the request / response data based on being requested as an argument in your api_function. These apply as input parameters only, and can not be applied currently as output formats or transformations.
@hug.get()
def test_time(hug_timer):
return {'time_taken': float(hug_timer)}Directives may be accessed via an argument with a hug_ prefix, or by using Python 3 type annotations. The latter is the more modern approach, and is recommended. Directives declared in a module can be accessed by using their fully qualified name as the type annotation (ex: module.directive_name).
Aside from the obvious input transformation use case, directives can be used to pipe data into your API functions, even if they are not present in the request query string, POST body, etc. For an example of how to use directives in this way, see the authentication example in the examples folder.
Adding your own directives is straight forward:
@hug.directive()
def square(value=1, **kwargs):
'''Returns passed in parameter multiplied by itself'''
return value * value
@hug.get()
@hug.local()
def tester(value: square=10):
return value
tester() == 100For completeness, here is an example of accessing the directive via the magic name approach:
@hug.directive()
def multiply(value=1, **kwargs):
'''Returns passed in parameter multiplied by itself'''
return value * value
@hug.get()
@hug.local()
def tester(hug_multiply=10):
return hug_multiply
tester() == 100Output Formatters a function that takes the output of your API function and formats it for transport to the user of the API.
@hug.default_output_format()
def my_output_formatter(data):
return "STRING:{0}".format(data)
@hug.get(output=hug.output_format.json)
def hello():
return {'hello': 'world'}as shown, you can easily change the output format for both an entire API as well as an individual API call
Input Formatters a function that takes the body of data given from a user of your API and formats it for handling.
@hug.default_input_format("application/json")
def my_input_formatter(data):
return ('Results', hug.input_format.json(data))Input formatters are mapped based on the content_type of the request data, and only perform basic parsing. More detailed parsing should be done by the Type Annotations present on your api_function
Middleware functions that get called for every request a hug API processes
@hug.request_middleware()
def process_data(request, response):
request.env['SERVER_NAME'] = 'changed'
@hug.response_middleware()
def process_data(request, response, resource):
response.set_header('MyHeader', 'Value')You can also easily add any Falcon style middleware using:
__hug__.http.add_middleware(MiddlewareObject())Parameter mapping can be used to override inferred parameter names, eg. for reserved keywords:
import marshmallow.fields as fields
...
@hug.get('/foo', map_params={'from': 'from_date'}) # API call uses 'from'
def get_foo_by_date(from_date: fields.DateTime()):
return find_foo(from_date)Input formatters are mapped based on the content_type of the request data, and only perform basic parsing. More detailed parsing should be done by the Type Annotations present on your api_function
hug enables you to organize large projects in any manner you see fit. You can import any module that contains hug decorated functions (request handling, directives, type handlers, etc) and extend your base API with that module.
For example:
something.py
import hug
@hug.get('/')
def say_hi():
return 'hello from something'Can be imported into the main API file:
__init__.py
import hug
from . import something
@hug.get('/')
def say_hi():
return "Hi from root"
@hug.extend_api('/something')
def something_api():
return [something]Or alternatively - for cases like this - where only one module is being included per a URL route:
#alternatively
hug.API(__name__).extend(something, '/something')By default, hug returns an auto generated API spec when a user tries to access an endpoint that isn't defined. If you would not like to return this spec you can turn off 404 documentation:
From the command line application:
hug -nd -f {file} #nd flag tells hug not to generate documentation on 404Additionally, you can easily create a custom 404 handler using the hug.not_found decorator:
@hug.not_found()
def not_found_handler():
return "Not Found"This decorator works in the same manner as the hug HTTP method decorators, and is even version aware:
@hug.not_found(versions=1)
def not_found_handler():
return ""
@hug.not_found(versions=2)
def not_found_handler():
return "Not Found"When using the get and cli method decorator on coroutines, hug will schedule
the execution of the coroutine.
Using asyncio coroutine decorator
@hug.get()
@asyncio.coroutine
def hello_world():
return "Hello"Using Python 3.5 async keyword.
@hug.get()
async def hello_world():
return "Hello"NOTE: Hug is running on top Falcon which is not an asynchronous server. Even if using asyncio, requests will still be processed synchronously.
If you like to develop in Docker and keep your system clean, you can do that but you'll need to first install Docker Compose.
Once you've done that, you'll need to cd into the docker directory and run the web server (Gunicorn) specified in ./docker/gunicorn/Dockerfile, after which you can preview the output of your API in the browser on your host machine.
$ cd ./docker
# This will run Gunicorn on port 8000 of the Docker container.
$ docker-compose up gunicorn
# From the host machine, find your Dockers IP address.
# For Windows & Mac:
$ docker-machine ip default
# For Linux:
$ ifconfig docker0 | grep 'inet' | cut -d: -f2 | awk '{ print $1}' | head -n1By default, the IP is 172.17.0.1. Assuming that's the IP you see, as well, you would then go to http://172.17.0.1:8000/ in your browser to view your API.
You can also log into a Docker container that you can consider your work space. This workspace has Python and Pip installed so you can use those tools within Docker. If you need to test the CLI interface, for example, you would use this.
$ docker-compose run workspace bashOn your Docker workspace container, the ./docker/templates directory on your host computer is mounted to /src in the Docker container. This is specified under services > app of ./docker/docker-compose.yml.
bash-4.3# cd /src
bash-4.3# tree
.
├── __init__.py
└── handlers
├── birthday.py
└── hello.py
1 directory, 3 fileshug takes security and quality seriously. This focus is why we depend only on thoroughly tested components and utilize static analysis tools (such as bandit and safety) to verify the security of our code base. If you find or encounter any potential security issues, please let us know right away so we can resolve them.
To report a security vulnerability, please use the Tidelift security contact. Tidelift will coordinate the fix and disclosure.
HUG simply stands for Hopefully Useful Guide. This represents the project's goal to help guide developers into creating well written and intuitive APIs.
Thanks and I hope you find this hug helpful as you develop your next Python API!
~Timothy Crosley
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent