Clients originally provided requests wrappers to encourage best practices, particularly always using Sessions to connect to the same host or api endpoint. The primary goals were:

• provide a Client object with a convenient constructor
• support a base url so that requests can provide a relative path
• provide the same interface for asyncio

Since then httpx has emerged as the successor to requests, and supports the above features natively. So clients.Client can be replaced with httpx.Client for most use cases. The project will continue to be maintained for additional features, such as the Resource object.

Typical requests usage is redundant and inefficient, by not taking advantage of connection pooling.

r = requests.get('https://api.github.com/user', headers={'authorization': token})
r = requests.get('https://api.github.com/user/repos', headers={'authorization': token})

Using sessions is the better approach, but more verbose and in practice requires manual url joining.

s = requests.Session()
s.headers['authorization'] = token
r = s.get('https://api.github.com/user')
r = s.get('https://api.github.com/user/repos')

Clients make using sessions easier, with implicit url joining.

client = clients.Client('https://api.github.com/', headers={'authorization': token})
r = client.get('user')
r = client.get('user/repos')

Resources extend Clients to implicitly handle response content, with proper checking of status_code and content-type.

github = clients.Resource('https://api.github.com/', headers={'authorization': token})
for repo in github.get('user/repos', params={'visibility': 'public'}):
    ...

Resources also implement syntactic support for methods such as getattr and call, providing most of the benefits of custom clients as is.

for repo in github.user.repos(visibility='public'):
    ...

Asynchronous variants of all client types are provided, e.g., AsyncClient. Additional clients for RPC, GraphQL, and proxies also provided.

% pip install clients

• httpx

100% branch coverage.

% pytest [--cov]