feat:declare optional schema on streamsync state

[FabienArcellier]

implement #277

Schema declaration on the Application state allows Streamsync to handle complex serialization scenario and empower your IDE and toolchains to provide autocomplete and type checking.

I will create another PR to handle compute property.

---

This PR fix #226

without schema, the button is working only one time

with schema

---

---

[FabienArcellier]

Rewriting a module variable there introduces a risk of introducing futur bugs with current accessors. When a variable is import from external module, its value is freezed. There is a risk to replace the value of the variable without updating external module.

To reduce the risk of issue, I would advocate to expose a function that return the variable initial_state(). Using it, we remove the risk of freezing the value.

[ramedina86]

So, keeping the global variable but using an accessor rather than importing it directly?

Like this?

def get_initial_state():
  return initial_state

[FabienArcellier]

yes exactly

[ramedina86]

There are some minor comments about style, e.g. British spelling, capitalisation of some words, etc, but I'll take care of them. Nice docs though!

[ramedina86]

What is this? Is this necessary or a separate improvement to make it look more like a Dict?

[FabienArcellier]

I need a way to iterate through keys in state proxy in the test test_replace_dictionary_content_in_a_state_with_schema_should_transform_it_in_state_proxy_and_trigger_mutation

items = _state['items']
items = {k: v for k, v in items.items() if k != "Apple"}
_state["items"] = items

I may remove this method and doing something like that :

items = _state['items']
items = {k: v for k, v in items..to_dict().items() if k != "Apple"}
_state["items"] = items

[mmikita95]

Potentially reintroduces #171; clarification needed.

[mmikita95]

There's a problem with this: it counts 'initial' mutations (i.e. first call to the get_mutations_as_dict) well, but can't track them after it.

Consider this example:

# When
items = _state['items']
m = _state._state_proxy.get_mutations_as_dict()
# Initial mutation
assert m == {
    '+items': None,
    '+items.Apple': None,
    '+items.Apple.name': 'Apple',
    '+items.Apple.type': 'fruit',
    '+items.Cucumber': None,
    '+items.Cucumber.name': 'Cucumber',
    '+items.Cucumber.type': 'vegetable',
    '+items.Lettuce': None,
    '+items.Lettuce.name': 'Lettuce',
    '+items.Lettuce.type': 'vegetable'
} # True
items = {k: v for k, v in items.items() if k != "Apple"}
_state["items"] = items

# Then
m = _state._state_proxy.get_mutations_as_dict()
# equals to {'+items': None}
# which will nullify the "items" on frontend
# should have been {'-items.Apple': None}
# or {
#    '+items': None,
#    '+items.Cucumber': None,
#    '+items.Cucumber.name': 'Cucumber',
#    '+items.Cucumber.type': 'vegetable',
#    '+items.Lettuce': None,
#    '+items.Lettuce.name': 'Lettuce',
#    '+items.Lettuce.type': 'vegetable'
#} 

Our biggest challenge with this are nested structures (lists and dicts), as we can only see the upper level.

[FabienArcellier]

Thanks @mmikita95. There was something wrong. I set some changes trying to fix it. Could you take a look ?

[mmikita95]

In regard to the previous comment - these StateProxy conversions/reinitializations were in place to avoid this kind of issue (though, not accounting for nested lists). Could you give a little more insight on motivation behind removing this, so I could better understand on how we should proceed?

[FabienArcellier]

I have to move StateProxy creation in the State because, it's this object that know the schema. You should take a look on the method State._set_state_item.

[raaymax]

I see that tests for state propagation proven to be useful :D

    Expected string: "{\"a\":1,\"b\":2,\"c\":{\"d\":3}}"
    Received string: "{\"a\":1,\"b\":2,\"c\":null}"