aio-request 0.2.0a4

Various strategies for sending requests

aio-request

This library simplifies an interaction between microservices:

1. Allows sending requests using various strategies
2. Propagates a deadline and a priority of requests
3. Exposes client/server metrics

Example:

import aiohttp
import aio_request

async with aiohttp.ClientSession() as client_session:
    client = aio_request.setup(
        transport=aio_request.AioHttpTransport(client_session),
        endpoint="http://endpoint:8080/",
    )
    response_ctx = client.request(
        aio_request.get("thing"),
        deadline=aio_request.Deadline.from_timeout(5)
    )
    async with response_ctx as response:
        pass  # process response here

Request strategies

The following strategies are supported:

1. Single attempt. Only one attempt is sent.
2. Sequential. Attempts are sent sequentially with delays between them.
3. Parallel. Attempts are sent in parallel one by one with delays between them.

Attempts count and delays are configurable.

Example:

import aiohttp
import aio_request

async with aiohttp.ClientSession() as client_session:
    client = aio_request.setup(
        transport=aio_request.AioHttpTransport(client_session),
        endpoint="http://endpoint:8080/",
    )
    response_ctx = client.request(
        aio_request.get("thing"),
        deadline=aio_request.Deadline.from_timeout(5),
        strategy=aio_request.parallel_strategy(
            attempts_count=3,
            delays_provider=aio_request.linear_backoff_delays(min_delay_seconds=0.1, delay_multiplier=0.1)
        )
    )
    async with response_ctx as response:
        pass  # process response here

Deadline & priority propagation

To enable it for the server side a middleware should be configured:

import aiohttp.web
import aio_request

app = aiohttp.web.Application(middlewares=[aio_request.aiohttp_middleware_factory()])

Expose client/server metrics

To enable client metrics, just install prometheus-client

import aiohttp
import aio_request

async with aiohttp.ClientSession() as client_session:
    client = aio_request.setup(
        transport=aio_request.AioHttpTransport(
            client_session,
        ),
        endpoint="http://endpoint:8080/",
    )

It is an example of how it should be done for aiohttp and prometheus.

To enable client metrics just install prometheus-client and use the following code:

import aiohttp.web
import aio_request

app = aiohttp.web.Application(
    middlewares=[
        aio_request.aiohttp_middleware_factory()
    ]
)

Circuit breaker

import aiohttp
import aio_request

async with aiohttp.ClientSession() as client_session:
    client = aio_request.setup_v2(
        transport=aio_request.AioHttpTransport(client_session),
        endpoint="http://endpoint:8080/",
        circuit_breaker=aio_request.DefaultCircuitBreaker[str, int](
            break_duration=1.0,
            sampling_duration=1.0,
            minimum_throughput=2,
            failure_threshold=0.5,
        ),
    )

In the case of requests count >= minimum throughput(>=2) in sampling period(1 second) the circuit breaker will open if failed requests count/total requests count >= failure threshold(50%).

unreleased

• Support httpx transport
• Drop python 3.9/3.10 support, support only 3.11/3.12/3.13. Related PRs: #222, #266, #275
• Deprecation of MetricsProvider interface. For the backward compatibility, prometheus-client is conditionally imported. To use it, install prometheus-client. Related PRs: #271, #218, #268
• Removal of unused Client interface
• Do not retry low timeout response
• Merge setup functions

v0.1.34 (2024-11-05)

• Try to get metrics provider from transport in setup_v2 if no metrics provider is passed

v0.1.33 (2024-10-29)

• Only yarl >= 1.12 is supported

v0.1.32 (2024-10-18)

• Endpoint provider

v0.1.31 (2024-09-05)

• Only yarl < 1.9.10 is supported

v0.1.30 (2023-07-23)

• Removal of tracing support
• Drop python 3.8 support

v0.1.29 (2023-04-27)

• Stop losing redirects params in headers update

v0.1.28 (2023-04-27)

• Add allow_redirects and max_redirects options to request

v0.1.27 (2023-02-16)

• Maintenance release

v0.1.26 (2022-11-02)

• Add python 3.11 support

v0.1.25 (2022-08-25)

• Reverted: URL-encode path_parameters - let user decide what to do

v0.1.24 (2022-07-04)

• URL-encode path_parameters

v0.1.23 (2022-02-08)

• Reject throttling(too many requests) status code

v0.1.22 (2022-01-08)

• Return default json expected content_type to "application/json"
• Release aiohttp response instead of close
• Validate json content-type

v0.1.21 (2022-01-05)

• Content type should be None in Response.json()

v0.1.20 (2022-01-05)

• Do not expect json content type by default

v0.1.19 (2021-11-01)

• Support async-timeout 4.0+

v0.1.18 (2021-09-08)

• Reexport explicitly

v0.1.17 (2021-09-01)

• Fix patch/patch_json visibility

v0.1.16 (2021-09-01)

• Support patch method

v0.1.15 (2021-09-01)

• Clean up resources in single shield

v0.1.14 (2021-08-18)

• Keys should be materialized if dict is changed in loop

v0.1.13 (2021-08-15)

• Circuit breaker

v0.1.12 (2021-07-21)

• Basic repr implementation

v0.1.11 (2021-07-21)

• Fix Request.update_headers, add Request.extend_headers #59

v0.1.10 (2021-07-20)

• Add Response.is_json property to check whether content-type is json compatible #58
• Tracing support #54,
• Configuration of a new pipeline