Incrementally build iter actions list

[hasier]

First PR after breaking down #433

• Second PR Support async actions hasier/tenacity#1 Support async actions #437
• Third PR Add async strategies hasier/tenacity#2 Add async strategies #451

An attempt at DRYing the iter() function in order to support async callbacks.

The approach I have taken is to build a list of actions that iter() needs to go through, where each action is just a piece of the whole current logic, and each step can further extend the list of actions. 3 pieces have been taken out to redefine in specific Retrying implementations, i.e. asyncio. This way we can make sure we use coroutines for those calls, as well as dynamically wrapping each function as a coroutine, allowing for both sync and async strategies and callbacks.

[mergify]

⚠️ No release notes detected. Please make sure to use reno to add a changelog entry.

[jd]

Can we have something either more structured or split?
I am no fan of having a dict with Any and random keys TBH.

At least a TypedDict?

[hasier]

Sounds fair, wdyt? 538686b

tbh I just copied the approach from the statistics property, and I was tempted to change the local() implementation for a ContextVar and maybe introducing dataclasses, but seemed like a bit too much to do together with these changes. Maybe something to tackle separately if that's something you'd like to see.

[hasier]

@jd I noticed black checks are also failing in CI, but I think there is an incompatibility between black and pep8. The first one is trying to remove some whitespace, whereas the latter complains it wants the whitespace back. Is this something you can look at in the main branch maybe?

[jd]

I switched the project to ruff to solve this.

[hasier]

I switched the project to ruff to solve this.

Thanks @jd! Just re-formatted the changes with ruff 😊

[mergify]

⚠️ No release notes detected. Please make sure to use reno to add a changelog entry.

Pull request has been modified.

[hasier]

Sorry @jd just saw that dataclass(slots=True) is only available from Python 3.10 onwards 🤦 I just fixed it by conditionally adding it, hopefully that looks ok to you! 5b11340

[hasier]

@jd how would you want to go about the chain of PRs? Do you want me to add a reno note here and merge this set of changes first, or would you prefer to merge the 2 other PRs here, then just merge 1 big PR to main? Happy to follow whichever process you prefer, let me know if you want me to update the PRs in any specific way.

[jd]

As long as there is no external changes, we can skip the changelog entry.