Quickstart¶
Ready to write your first API client with Uplink? This guide will walk you through what you’ll need to know to get started.
First, make sure you’ve installed (or updated) Uplink:
$ pip install -U uplink
Defining an API Client¶
Writing a structured API client with Uplink is very simple.
To start, create a subclass of Consumer
. For example,
here’s the beginning of our GitHub client (we’ll add some methods to
this class soon):
from uplink import Consumer
class GitHub(Consumer):
...
When creating an instance of this consumer, we can use the base_url
constructor argument to identify the target service. In our case, it’s
GitHub’s public API:
github = GitHub(base_url="https://api.github.com/")
Note
base_url
is especially useful for creating clients that
target separate services with similar APIs; for example, we could use
this GitHub consumer to also create clients for any GitHub
Enterprise instance for projects hosted outside of the public
GitHub.com service. Another example is
creating separate clients for a company’s production and staging
environments, which are typically hosted on separate domains but
expose the same API.
So far, this class looks like any other Python class. The real magic happens when you define methods to interact with the webservice using Uplink’s HTTP method decorators, which we cover next.
Making a Request¶
With Uplink, making a request to a webservice is as simple as invoking a method.
Any method of a Consumer
subclass can be
decorated with one of Uplink’s HTTP method decorators:
get
, post
, put
,
patch
, head
, and delete
:
class GitHub(Consumer):
@get("repositories")
def get_repos(self):
"""List all public repositories."""
As shown above, the method’s body can be left empty.
The decorator’s first argument is the resource endpoint (this
is the relative URL path from base_url
, which we covered above):
@get("repositories")
You can also specify query parameters:
@get("repositories?since=364")
Finally, invoke the method to send a request:
>>> github = GitHub(base_url="https://api.github.com/")
>>> github.get_repos()
<Response [200]>
>>> _.url
https://api.github.com/repositories
By default, uplink uses Requests, so the response we get back
from GitHub is wrapped inside a requests.Response
instance. (If
you want, you can swap out
Requests for a different backing HTTP client, such as aiohttp.)
URL Manipulation¶
Resource endpoints can include URI template parameters that depend on method
arguments. A simple URI parameter is an alphanumeric string surrounded
by {
and }
.
To match the parameter with a method argument, either match the argument’s name with the alphanumeric string, like so:
@get("users/{username}")
def get_user(self, username): pass
or use the Path
annotation.
@get("users/{username}")
def get_user(self, name: Path("username")): pass
Query
parameters can also be added dynamically
by method arguments.
@get("users/{username}/repos")
def get_repos(self, username, sort: Query): pass
For “catch-all” or complex query parameter combinations, a
QueryMap
can be used:
@get("users/{username}/repos")
def get_repos(self, username, **options: QueryMap): pass
You can set static query parameters for a method using the
params
decorator.
@params({"client_id": "my-client", "client_secret": "****"})
@get("users/{username}")
def get_user(self, username): pass
params
can be used as a class decorator for query
parameters that need to be included with every request:
@params({"client_id": "my-client", "client_secret": "****"})
class GitHub(Consumer):
...
Header Manipulation¶
You can set static headers for a method using the headers
decorator.
@headers({
"Accept": "application/vnd.github.v3.full+json",
"User-Agent": "Uplink-Sample-App"
})
@get("users/{username}")
def get_user(self, username): pass
headers
can be used as a class decorator for headers that
need to be added to every request:
@headers({
"Accept": "application/vnd.github.v3.full+json",
"User-Agent": "Uplink-Sample-App"
})
class GitHub(Consumer):
...
A request header can depend on the value of a method argument by using
the Header
function parameter annotation:
@get("user")
def get_user(self, authorization: Header("Authorization"):
"""Get an authenticated user."""
Request Body¶
The Body
annotation identifies a method argument as the
the HTTP request body:
@post("user/repos")
def create_repo(self, repo: Body): pass
This annotation works well with the keyword arguments parameter (denoted
by the **
prefix):
@post("user/repos")
def create_repo(self, **repo_info: Body): pass
Moreover, this annotation is useful when using supported serialization formats, such as JSON and Protocol Buffers. Take a look at this guide for more about serialization with Uplink.
Form Encoded, Multipart, and JSON Requests¶
Methods can also be declared to send form-encoded, multipart, and JSON data.
Form-encoded data is sent when form_url_encoded
decorates
the method. Each key-value pair is annotated with a Field
annotation:
@form_url_encoded
@patch("user")
def update_user(self, name: Field, email: Field): pass
Multipart requests are used when multipart
decorates the
method. Parts are declared using the Part
annotation:
@multipart
@put("user/photo")
def upload_photo(self, photo: Part, description: Part): pass
JSON data is sent when json
decorates the method. The
Body
annotation declares the JSON payload:
@json
@patch("user")
def update_user(self, **user_info: uplink.Body):
"""Update an authenticated user."""
Alternatively, the Field
annotation declares a JSON
field:
@json
@patch("user")
def update_user_bio(self, bio: Field):
"""Update the authenticated user's profile bio."""
Handling JSON Responses¶
Many modern public APIs serve JSON responses to their clients.
If your Consumer
subclass accesses a JSON API, you can
decorate any method with returns.json
to
directly return the JSON response, instead of a response object, when
invoked:
class GitHub(Consumer):
@returns.json
@get("users/{username}")
def get_user(self, username):
"""Get a single user."""
>>> github = GitHub("https://api.github.com")
>>> github.get_user("prkumar")
{'login': 'prkumar', 'id': 10181244, ...
You can also target a specific field of the JSON response by using the
decorator’s member
argument to select the target JSON field name:
class GitHub(Consumer):
@returns.json(member="blog")
@get("users/{username}")
def get_blog_url(self, username):
"""Get the user's blog URL."""
>>> github.get_blog_url("prkumar")
"https://prkumar.io"
Note
JSON responses may represent existing Python classes in your
application (for example, a GitHubUser
). Uplink supports this
kind of conversion (i.e., deserialization), and we detail this
support in the next guide.
Persistence Across Requests from a Consumer
¶
The session
property of a Consumer
instance exposes
the instance’s configuration and allows for the persistence of certain
properties across requests sent from that instance.
You can provide default headers and query parameters for requests sent from a
consumer instance through its session
property, like so:
class GitHub(Consumer):
def __init__(self, username, password)
# Creates the API token for this user
api_key = create_api_key(username, password)
# Send the API token as a query parameter with each request.
self.session.params["api_key"] = api_key
@get("user/repos")
def get_user_repos(self, sort_by: Query("sort")):
"""Lists public repositories for the authenticated user."""
Headers and query parameters added through the session
are
applied to all requests sent from the consumer instance.
github = GitHub("prkumar", "****")
# Both `api_key` and `sort` are sent with the request.
github.get_user_repos(sort_by="created")
Notably, in case of conflicts, the method-level headers and parameters override the session-level, but the method-level properties are not persisted across requests.
Response and Error Handling¶
Sometimes, you need to validate a response before it is returned or even calculate a new return value from the response. Or, you may need to handle errors from the underlying client before they reach your users.
With Uplink, you can address these concerns by registering a callback
with one of these decorators: response_handler
and
error_handler
.
response_handler
registers a callback to intercept
responses before they are returned (or deserialized):
def raise_for_status(response):
"""Checks whether or not the response was successful."""
if 200 <= response.status <= 299:
raise UnsuccessfulRequest(response.url)
# Pass through the response.
return response
class GitHub(Consumer):
@response_handler(raise_for_status)
@post("user/repo")
def create_repo(self, name: Field):
"""Create a new repository."""
error_handler
registers a callback to handle an
exception thrown by the underlying HTTP client
(e.g., requests.Timeout
):
def raise_api_error(exc_type, exc_val, exc_tb):
"""Wraps client error with custom API error"""
raise MyApiError(exc_val)
class GitHub(Consumer):
@error_handler(raise_api_error)
@post("user/repo")
def create_repo(self, name: Field):
"""Create a new repository."""
To apply a callback onto all methods of a Consumer
subclass, you can simply decorate the class itself:
@error_handler(raise_api_error)
class GitHub(Consumer):
...
Notably, the decorators can be stacked on top of one another to chain their behaviors:
@response_handler(check_expected_headers) # Second, check headers
@response_handler(raise_for_status) # First, check success
class TodoApp(Consumer):
...