[alexanto]

async ounterpart of #473

[ppicom]

Hi there! I haven't looked at the code, just at the number of LoC and files that changed. Have we considered splitting this PR in smaller ones? I know it's the top of the stack, a lot needs to happen for this one to get merged. But just throwing it out there to start thinking about whether it's possible

[Copilot]

🟡 Changes recommended

Async transport validation rejects valid coroutine-returning methods wrapped by synchronous decorators, causing AsyncUnleashClient construction to fail before any request.

1 open issue

• Accept decorated async transport methods during validation · New

What changed in this PR

Extends public transport customization to AsyncUnleashClient, matching the synchronous client’s support for user-managed sessions and alternative HTTP libraries.

Changes:

• Exposes async transports and adds constructor validation.
• Adds custom aiohttp sessions, headers, and explicit resource ownership.
• Expands tests, documentation, and examples for async transport usage.

File	Description
UnleashClient/​transport.py	Shares transport validation and authorization-header checks.
UnleashClient/​clients/​unleash_client.py	Uses shared transport validation.
UnleashClient/​clients/​async_unleash_client.py	Accepts and validates custom async transports.
UnleashClient/​_async_unleash_api_client.py	Looks up response ETags case-insensitively.
UnleashClient/​_aiohttp_transport.py	Supports user sessions, custom headers, and ownership-aware cleanup.
UnleashClient/​__init__.py	Exports async transport interfaces and implementation.
tests/​unit_tests/​test_aiohttp_transport.py	Tests session ownership, headers, and response handling.
tests/​unit_tests/​clients/​test_async_unleash_client_transport.py	Tests transport integration, validation, and examples.
README.md	Documents async transport configuration and lifecycle.
examples/​httpx_async_transport.py	Demonstrates an httpx-backed async transport.
examples/​fake_async_transport.py	Demonstrates async testing without a server.
CHANGELOG.md	Records public async transport support.

_{Review effort: Balanced}

[alexanto]

Hi there! I haven't looked at the code, just at the number of LoC and files that changed. Have we considered splitting this PR in smaller ones? I know it's the top of the stack, a lot needs to happen for this one to get merged. But just throwing it out there to start thinking about whether it's possible

i rebased it so it appears much smaller and the actual code is change is ~200 lines, the rest is just tests and changelog, so i hope it is not that horrible now

[sighphyre]

I think this has come together nicely, I think we should rethink the check transport functions responsibilities and coupling a little but other than that, it looks sound

[sighphyre]

I think this beast is now too big. Some thoughts here:

1. This is doing 3 things, first it's checking requests stuff, second its checking async stuff if it was called by the async paths, third it should be checking sync stuff if it was called by the sync paths
2. I think it's actually lost the sync check, from the previous PR
3. This is aggressively tying 3 unrelated sections of the application together - the requests import, the sync client validation and the async client validation
4. 5 parameters is leaning on the side of too many parameters, especially when we pass boolean flags to target behavior is a sign that the function should be multiple functions

Sooo... after that whole essay, I would suggest not moving this here. Leave this function in the sync client, don't care what you call it. Have that function check that we have an synchronous execute method. If you really want, include the check for requests. Then make another function in the async client, can be the same name, don't care. Have that function check that we have a async execute - we can leave the requests check off that probably

Think it's one of those places where a little copying is better than a lot of coupling

[alexanto]

yes this got a little messy, split it up now, kept the name