popget

A simple no-bells-and-whistles REST-API client, optionally supporting async requests.

We use this for service-to-service requests in our heterogenous microservices environment.

Usage

pip install popget

Configuration

Settings are intended to be configured primarily via a python file, such as your existing Django settings.py. To bootstrap this, there are a couple of env vars to control how config is loaded:

See source of popget/conf/settings.py for more details.

Some useful config keys (all of which are prefixed with POPGET_ by default):

• POPGET_CLIENT_DEFAULT_USER_AGENT when making requests, popget will use this string as the user agent.
• POPGET_CLIENT_DEFAULT_HEADERS when making requests, popget will add these headers by default to the request, but they can still be overridden when set explicitly.
• POPGET_CLIENT_TIMEOUT if None then no timeout, otherwise this timeout (in seconds) will be applied to all requests. Requests which timeout will return a 504 response, which will be raised as an HTTPError.

APIClient

You will sub-class APIClient to make your API. You do not need to instantiate the client, all methods are class-methods.

eg

from popget import APIClient, Arg, GetEndpoint

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'

    get_things = GetEndpoint(
        '/things/{user_id}/',  # url format string
        querystring_args=(
            Arg('type', required=True),  # required querystring param (validated on call)
        )
    )

Results in a client method you can call like:

data = ThingServiceClient.get_things(user_id=2345, type='cat')

Which will perform a request like:

GET http://things.depop.com/things/2345/?type=cat

If response was "Content-Type: application/json" then data is already deserialized.

Under Python 3 there is a further distinction between str and bytes. If the Content-Type header contains text/ then the returned value will be encoded to str (by underlying python-requests lib). Other content types will return bytes.

We use raise_for_status so anything >= 400 will raise a requests.HTTPError.

APIEndpoint

APIEndpoint is the base class for endpoint methods. GetEndpoint, PostEndpoint, PutEndpoint, PatchEndpoint and DeleteEndpoint are provided for convenience, allowing to omit the method arg.

Params from url path (format string), querystring and request headers (format string of value portion) will be extracted and made available as kwargs on the resulting callable method on your client class.

This means arg names must be unique across all three sources of args. This is feasible because path and header args can be freely chosen when implementing the client (they are just format string identifiers rather than part of the REST API itself like querystring args are).

e.g.

from popget import APIClient, Arg, GetEndpoint

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'

    get_things = GetEndpoint(
        '/things/{user_id}/',  # url (format string)
        querystring_args=(
            Arg('type', required=True),
            Arg('offset_id'),
            Arg('limit', default=25),
        ),
        request_headers={      # added to all requests
            'Authorization': 'Bearer {access_token}'  # (format string)
        }
    )

This will give you a client with a get_things method you can call like:

response_data = ThingServiceClient.get_things(
    user_id=123,
    type='cat',
    offset_id='65345ff34e344ab53c',
    limit=20,
    access_token='87a64c98b62d39e8625f',
)

Querystring args can have a callable as the default value, e.g.:

from datetime import datetime
from popget import APIClient, Arg, GetEndpoint

def now():
    return datetime.now().isoformat()

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'

    get_things = GetEndpoint(
        '/things/{user_id}/',  # url (format string)
        querystring_args=(
            Arg('since', default=now),
        ),
        request_headers={      # added to all requests
            'Authorization': 'Bearer {access_token}'  # (format string)
        }
    )

response_data = ThingServiceClient.get_things(user_id=123)
# GET http://things.depop.com/things/123/?since=2018-02-09T13:31:10.569481

You can still pass extra args down into the requests lib on a per-call basis by using _request_kwargs:

response_data = ThingServiceClient.get_things(
    user_id=123,
    type='cat',
    offset_id='65345ff34e344ab53c',
    limit=20,
    access_token='87a64c98b62d39e8625f',
    _request_kwargs={
        'headers': {
            'X-Depop-WTF': 'something something'
        }
    },
)

And for calls with a request body:

from popget import APIClient, PostEndpoint, BodyType

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'

    new_thing = PostEndpoint(
        '/things/',
        body_required=True,
        body_type=BodyType.FORM_ENCODED,  # or BodyType.JSON - sets content-type header
        request_headers={
            'Authorization': 'Bearer {access_token}',
        }
    )

response_data = ThingServiceClient.new_thing(
    access_token='87a64c98b62d39e8625f',
    body={
        'type': 'dog',
        'name': 'fido',
    }
)

You can also pass a custom requests.Session instance on a per-request basis using the _session kwarg:

from django.conf import settings
from requests_oauthlib import OAuth1Session

from myproject.twitter.client import TwitterClient

def tweet_from(user, message, **extra):
    oauth_session = OAuth1Session(
        settings.TWITTER_CONSUMER_KEY,
        client_secret=settings.TWITTER_CONSUMER_SECRET,
        resource_owner_key=user.tw_access_token,
        resource_owner_secret=user.tw_access_token_secret,
    )
    body = {
        'status': message,
    }
    body.update(extra)
    return TwitterClient.update_status(body=body, _session=oauth_session)

Asynchronous

Optional support for asynchronous requests is provided, via a requests-futures backend.

pip install popget[threadpool]

An async variant of the APIClient is provided which will have both async and blocking versions of all endpoint methods.

from popget import Arg, GetEndpoint
from popget.async.threadpool import APIClient
import requests

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'

    get_things = GetEndpoint(
        '/things/{user_id}/',            # url format string
        querystring_args=(
            Arg('type', required=True),  # required querystring param (validated on call)
        ),
    )

# blocking:
data = ThingServiceClient.get_things(user_id=2345, type='cat')

# async:
future = ThingServiceClient.async_get_things(user_id=2345, type='cat')
# response is parsed and may raise, as for blocking requests
try:
    data = future.result()
except requests.exceptions.HTTPError as e:
    print(repr(e))

The async endpoint methods will return a standard concurrent.futures.Future object.

See Python docs.

You can customise the name of the generated async endpoint methods:

class ThingServiceClient(APIClient):

    class Config:
        base_url = 'http://things.depop.com'
        async_method_template = '{}__async'

    get_things = GetEndpoint(
        '/things/{user_id}/',            # url format string
        querystring_args=(
            Arg('type', required=True),  # required querystring param (validated on call)
        ),
    )

future = ThingServiceClient.get_things__async(user_id=2345, type='cat')

Compatibility

This project is tested against:

Python 3.11

Running the tests

py.test (single python version)

It's also possible to run the tests locally, allowing for debugging of errors that occur.

Decide which Python version you want to test and create a virtualenv:

python -m virtualenv .venv -p python3.11
pip install -r requirements-test.txt
py.test -v -s tests/