Separate the high level client implementation from the underlying CRUD.
Project description
Python API Client
A client for communicating with an api should be a clean abstraction over the third part api you are communicating with. It should be easy to understand and have the sole responsibility of calling the endpoints and returning data.
To achieve this, APIClient
takes care of the other (often duplicated)
responsibilities, such as authentication and response handling, moving
that code away from the clean abstraction you have designed.
Quick links
- Installation
- Client in action
- Adding retries to requests
- Working with paginated responses
- Authenticating your requests
- Handling the formats of your responses
- Correctly encoding your outbound request data
- Handling bad requests and responses
- Endpoints as code
- Marshalling requests/responses
Installation
pip install api-client
Usage
Simple Example
from apiclient import APIClient
class MyClient(APIClient):
def list_customers(self):
url = "http://example.com/customers"
return self.get(url)
def add_customer(self, customer_info):
url = "http://example.com/customers"
return self.post(url, data=customer_info)
>>> client = MyClient()
>>> client.add_customer({"name": "John Smith", "age": 28})
>>> client.list_customers()
[
...,
{"name": "John Smith", "age": 28},
]
The APIClient
exposes a number of predefined methods that you can call
This example uses get
to perform a GET request on an endpoint.
Other methods include: post
, put
, patch
and delete
. More
information on these methods is documented in the Interface.
For a more complex use case example, see: Extended example
Retrying
To add some robustness to your client, the power of tenacity
has been harnessed to add a @retry_request
decorator to the apiclient
toolkit.
This will retry any request which responds with a 5xx status_code (which is normally safe
to do as this indicates something went wrong when trying to make the request), or when an
UnexpectedError
occurs when attempting to establish the connection.
@retry_request
has been configured to retry for a maximum of 5 minutes, with an exponential
backoff strategy. For more complicated uses, the user can use tenacity themselves to create
their own custom decorator.
Usage:
from apiclient import retry_request
class MyClient(APIClient):
@retry_request
def retry_enabled_method():
...
For more complex use cases, you can build your own retry decorator using tenacity along with the custom retry strategy.
For example, you can build a retry decorator that retries APIRequestError
which waits for 2 seconds between retries and gives up after 5 attempts.
import tenacity
from apiclient.retrying import retry_if_api_request_error
retry_decorator = tenacity.retry(
retry=retry_if_api_request_error(),
wait=tenacity.wait_fixed(2),
stop=tenacity.stop_after_attempt(5),
reraise=True,
)
Or you can build a decorator that will retry only on specific status codes (following a failure).
retry_decorator = tenacity.retry(
retry=retry_if_api_request_error(status_codes=[500, 501, 503]),
wait=tenacity.wait_fixed(2),
stop=tenacity.stop_after_attempt(5),
reraise=True,
)
Pagination
In order to support contacting pages that respond with multiple pages of data when making get requests,
add a @paginated
decorator to your client method. @paginated
can paginate the requests either where
the pages are specified in the query parameters, or by modifying the url.
Usage is simple in both cases; paginator decorators take a Callable with two required arguments:
by_query_params
-> callable takesresponse
andprevious_page_params
.by_url
-> callable takesrespones
andprevious_page_url
. The callable will need to return either the params in the case ofby_query_params
, or a new url in the case ofby_url
. If the response is the last page, the function should return None.
Usage:
from apiclient import paginated
def next_page_by_params(response, previous_page_params):
# Function reads the response data and returns the query param
# that tells the next request to go to.
return {"next": response["pages"]["next"]
def next_page_by_url(response, previous_page_url):
# Function reads the response and returns the url as string
# where the next page of data lives.
return response["pages"]["next"]["url"]
class MyClient(APIClient):
@paginated(by_query_params=next_page_by_params)
def paginated_example_one():
...
@paginated(by_url=next_page_by_url)
def paginated_example_two():
...
Authentication Methods
Authentication methods provide a way in which you can customize the client with various authentication schemes through dependency injection, meaning you can change the behaviour of the client without changing the underlying implementation.
The apiclient supports the following authentication methods, by specifying the initialized class on initialization of the client, as follows:
client = ClientImplementation(
authentication_method=<AuthenticationMethodClass>(),
response_handler=...,
request_formatter=...,
)
NoAuthentication
This authentication method simply does not add anything to the client, allowing the api to contact APIs that do not enforce any authentication.
Example:
client = ClientImplementation(
authentication_method=NoAuthentication(),
response_handler=...,
request_formatter=...,
)
QueryParameterAuthentication
This authentication method adds the relevant parameter and token to the client query parameters. Usage is as follows:
client = ClientImplementation(
authentication_method=QueryParameterAuthentication(parameter="apikey", token="secret_token"),
response_handler=...,
request_formatter=...,
)
Example. Contacting a url with the following data
http://api.example.com/users?age=27
Will add the authentication parameters to the outgoing request:
http://api.example.com/users?age=27&apikey=secret_token
HeaderAuthentication
This authentication method adds the relevant authorization header to the outgoing request. Usage is as follows:
client = ClientImplementation(
authentication_method=HeaderAuthentication(token="secret_value"),
response_handler=...,
request_formatter=...,
)
# Constructs request header:
{"Authorization": "Bearer secret_value"}
The Authorization
parameter and Bearer
scheme can be adjusted by
specifying on method initialization.
authentication_method=HeaderAuthentication(
token="secret_value"
parameter="apikey",
scheme="Token",
)
# Constructs request header:
{"apikey": "Token secret_value"}
Or alternatively, when APIs do not require a scheme to be set, you can specify it as a value that evaluates to False to remove the scheme from the header:
authentication_method=HeaderAuthentication(
token="secret_value"
parameter="token",
scheme=None,
)
# Constructs request header:
{"token": "secret_value"}
Additional header values can be passed in as a dict here when API's require more than one header to authenticate:
authentication_method=HeaderAuthentication(
token="secret_value"
parameter="token",
scheme=None,
extra={"more": "another_secret"}
)
# Constructs request header:
{"token": "secret_value", "more": "another_secret"}
BasicAuthentication
This authentication method enables specifying a username and password to APIs that require such.
client = ClientImplementation(
authentication_method=BasicAuthentication(username="foo", password="secret_value"),
response_handler=...,
request_formatter=...,
)
CookieAuthentication
This authentication method allows a user to specify a url which is used to authenticate an initial request, made at APIClient initialization, with the authorization tokens then persisted for the duration of the client instance in cookie storage.
These cookies use the http.cookiejar.CookieJar()
and are set on the
session so that all future requests contain these cookies.
As the method of authentication at the endpoint is not standardised
across API's, the authentication method can be customized using one of
the already defined authentication methods; QueryParameterAuthentication
,
HeaderAuthentication
, BasicAuthentication
.
client = ClientImplementation(
authentication_method=(
CookieAuthentication(
auth_url="https://example.com/authenticate",
authentication=HeaderAuthentication("1234-secret-key"),
),
response_handler=...,
request_formatter=...,
)
Response Handlers
Response handlers provide a standard way of handling the final response
following a successful request to the API. These must inherit from
BaseResponseHandler
and implement the get_request_data()
method which
will take the requests.Response
object and parse the data accordingly.
The apiclient supports the following response handlers, by specifying the class on initialization of the client as follows:
The response handler can be omitted, in which case no formatting is applied to the outgoing data.
client = ClientImplementation(
authentication_method=...,
response_handler=<ResponseHandlerClass>,
request_formatter=...,
)
RequestsResponseHandler
Handler that simply returns the original Response
object with no
alteration.
Example:
client = ClientImplementation(
authentication_method=...,
response_handler=RequestsResponseHandler,
request_formatter=...,
)
JsonResponseHandler
Handler that parses the response data to json
and returns the dictionary.
If an error occurs trying to parse to json then a UnexpectedError
will be raised.
Example:
client = ClientImplementation(
authentication_method=...,
response_handler=JsonResponseHandler,
request_formatter=...,
)
XmlResponseHandler
Handler that parses the response data to an xml.etree.ElementTree.Element
.
If an error occurs trying to parse to xml then a UnexpectedError
will be raised.
Example:
client = ClientImplementation(
authentication_method=...,
response_handler=XmlResponseHandler,
request_formatter=...,
)
YamlResponseHandler
Handler that parses the response data in yaml
format and returns the
dictionary. If an error occurs trying to parse the yaml then an UnexpectedError
will be raised.
Example:
client = ClientImplementation(
authentication_method=...,
response_handler=YamlResponseHandler,
request_formatter=...,
)
Request Formatters
Request formatters provide a way in which the outgoing request data can be encoded before being sent, and to set the headers appropriately.
These must inherit from BaseRequestFormatter
and implement the format()
method which will take the outgoing data
object and format accordingly
before making the request.
The apiclient supports the following request formatters, by specifying the class on initialization of the client as follows:
client = ClientImplementation(
authentication_method=...,
response_handler=...,
request_formatter=<RequestFormatterClass>,
)
JsonRequestFormatter
Formatter that converts the data into a json format and adds the
application/json
Content-type header to the outgoing requests.
Example:
client = ClientImplementation(
authentication_method=...,
response_handler=...,
request_formatter=JsonRequestFormatter,
)
Exceptions
The exception handling for api-client
has been designed in a way so that all exceptions inherit from
one base exception type: APIClientError
. From there, the exceptions have been broken down into the
following categories:
ResponseParseError
Something went wrong when trying to parse the successful response into the defined format. This could be due
to a misuse of the ResponseHandler, i.e. configuring the client with an XmlResponseHandler
instead of
a JsonResponseHandler
APIRequestError
Something went wrong when making the request. These are broken down further into the following categories to provide greater granularity and control.
RedirectionError
A redirection status code (3xx) was returned as a final code when making the request. This means that no data can be returned to the client as we could not find the requested resource as it had moved.
ClientError
A clienterror status code (4xx) was returned when contacting the API. The most common cause of these errors is misuse of the client, i.e. sending bad data to the API.
ServerError
The API was unreachable when making the request. I.e. a 5xx status code.
UnexpectedError
An unexpected error occurred when using the client. This will typically happen when attempting to make the request, for example, the client never receives a response. It can also occur to unexpected status codes (>= 600).
Endpoints
The apiclient also provides a convenient way of defining url endpoints with
use of the @endpoint
decorator. In order to decorate a class with @endpoint
the decorated class must define a base_url
attribute along with the required
resources. The decorator will combine the base_url with the resource.
Example:
from apiclient import endpoint
@endpoint(base_url="http://foo.com")
class Endpoint:
resource = "search"
>>> Endpoint.resource
"http://foo.com/search
Marshalling
The following decorators have been provided to marshal request data as python dataclasses to json and to unmarshal json directly into a python dataclass.
# Marshal dataclass -> json
@marshal_request(date_fmt: Optional[str] = None, datetime_fmt: Optional[str] = None)
# Unmarshal json -> dataclass
@unmarshal_response(schema: T, date_fmt: Optional[str] = None, datetime_fmt: Optional[str] = None)
Usage:
- Define the schema for your api in python dataclasses.
- Add the
@unmarshal_response
decorator to the api client method to transform the response directly into your defined schema.@unmarshal_response(List[Account]) def get_accounts(): ...
- Add the
@marshal_request
decorator to the api client method to translate the incoming dataclass into the required json for the endpoint:@marshal_request() def create_account(account: Account): ...
The marshalling functionality has been provided by: https://github.com/MikeWooster/jsonmarshal More usage examples can be found there.
Extended Example
from apiclient import APIClient, endpoint, paginated, retry_request
# Define endpoints, using the provided decorator.
@endpoint(base_url="https://jsonplaceholder.typicode.com")
class Endpoint:
todos = "todos"
todo = "todos/{id}"
def get_next_page(response):
return {
"limit": response["limit"],
"offset": response["offset"] + response["limit"],
}
# Extend the client for your API integration.
class JSONPlaceholderClient(APIClient):
@paginated(by_query_params=get_next_page)
def get_all_todos(self) -> dict:
return self.get(Endpoint.todos)
@retry_request
def get_todo(self, todo_id: int) -> dict:
url = Endpoint.todo.format(id=todo_id)
return self.get(url)
# Initialize the client with the correct authentication method,
# response handler and request formatter.
>>> client = JSONPlaceholderClient(
authentication_method=HeaderAuthentication(token="<secret_value>"),
response_handler=JsonResponseHandler,
request_formatter=JsonRequestFormatter,
)
# Call the client methods.
>>> client.get_all_todos()
[
{
'userId': 1,
'id': 1,
'title': 'delectus aut autem',
'completed': False
},
...,
{
'userId': 10,
'id': 200,
'title': 'ipsam aperiam voluptates qui',
'completed': False
}
]
>>> client.get_todo(45)
{
'userId': 3,
'id': 45,
'title': 'velit soluta adipisci molestias reiciendis harum',
'completed': False
}
# REST APIs correctly adhering to the status codes to provide meaningful
# responses will raise the appropriate exeptions.
>>> client.get_todo(450)
NotFound: 404 Error: Not Found for url: https://jsonplaceholder.typicode.com/todos/450
>>> try:
... client.get_todo(450)
... except APIClientError:
... print("All client exceptions inherit from APIClientError")
"All client exceptions inherit from APIClientError"
APIClient Interface
The APIClient
provides the following public interface:
-
post(self, endpoint: str, data: dict, params: OptionalDict = None)
Delegate to POST method to send data and return response from endpoint.
-
get(endpoint: str, params: OptionalDict = None)
Delegate to GET method to get response from endpoint.
-
put(endpoint: str, data: dict, params: OptionalDict = None)
Delegate to PUT method to send and overwrite data and return response from endpoint.
-
patch(endpoint: str, data: dict, params: OptionalDict = None)
Delegate to PATCH method to send and update data and return response from endpoint
-
delete(endpoint: str, params: OptionalDict = None)
Delegate to DELETE method to remove resource located at endpoint.
-
get_request_timeout() -> float
By default, all requests have been set to have a default timeout of 10.0 s. This is to avoid the request waiting forever for a response, and is recommended to always be set to a value in production applications. It is however possible to override this method to return the timeout required by your application.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
File details
Details for the file api-client-1.2.0.tar.gz
.
File metadata
- Download URL: api-client-1.2.0.tar.gz
- Upload date:
- Size: 18.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.2.0 pkginfo/1.5.0.1 requests/2.24.0 setuptools/41.2.0 requests-toolbelt/0.9.1 tqdm/4.47.0 CPython/3.8.3
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 2ab9e3a1a50e9b0ef14426bd1d008cbaf39c5844caf534cddad94af025d3945a |
|
MD5 | cee2ec7da1935484fde00131487f45d4 |
|
BLAKE2b-256 | 1daa35eecf11078fc47af36d80b299158f471597e983ffdbe0a08f9fd390ccf8 |