Add APIs for generic storage capability

[pksol]

Allow tests to store state data that can be used in the future:

• Add a "put" method which gets a key and data and stores that for the given test
• Add a "get" method which gets a key and returns the previously stored data for the given test
• Add a "state" attribute which tells the code if we're in recording (writing) mode or in fetching mode

This API is very generic and can be used to implement many capabilities.
For example, say I want to make sure my test performance doesn't degrade over time: I would store an initial benchmark in the recording phase and validate against that benchmark in future runs.

[pohmelie]

How this will looks like? Something like

if state is State.RECORDING:
    data = foo()
    put("data", data)
else:
    data = get("data")

I see many problems with such approach, and maybe this should be different lib, I'm not sure.

[pksol]

Yes, this is the basic idea I had.

It could be a separate lib since it's a more general usage than this library, maybe pytest-automock can even depend on that new library as the storage provider.

What are the problems you see with this approach?

[pohmelie]

The problem here, as I see it, that state can be used only in tests code itself, not in the code of library or project. Lets say I want to mock something with same mechanic as it is in automock. I need to bring some test-aware code into my library to dispatch by state. automock, instead do it implicitly, so your project code do not know is it a recording or is it a replaying mode.

[pksol]

I was thinking that when someone wants to use this capability, the code that touches state would be mostly in the fixtures, so individual tests are not aware of the state most of the time.