pandora

pandora copied to clipboard

An alternative approach to https://github.com/bitbrain/pandora/pull/162

The array serialization/deserialization is working in the test for Color.

Outstanding questions Should a reference deserialize into the entity itself or a reference type?

ie. What would make more sense?

get_array("Equipment") return Array[PandoraReference] and you manually .map(func(r): return r.get_entity()) or we do that step ourselves in the deserialization

get_array("Equipment") return Array[PandoraEntity] which feels a bit more logical to hide the reference.

Old save/load cycle

var old_color_array = old_entity_1.get_array("typed_array")
var new_color_array = new_entity_1.get_array("typed_array")

print("Got Array")
print(old_color_array)
print(new_color_array)

print(typeof(old_color_array[0]))
print(typeof(new_color_array[0]))

[(1, 0, 0, 1)]
["ff0000ff"]
20
4

New array code.

var old_color_array = old_entity_1.get_array("typed_array")
var new_color_array = new_entity_1.get_array("typed_array")

print("Got Array")
print(old_color_array)
print(new_color_array)

print(typeof(old_color_array[0]))
print(typeof(new_color_array[0]))

[(1, 0, 0, 1)]
[(1, 0, 0, 1)]
20
20

However, I am struggling a bit with the test because the minute I call either of

var old_color_array = old_entity_1.get_array("typed_array")
var new_color_array = new_entity_1.get_array("typed_array")

something inside of the ._entities dictionary is loading a recursive reference because running

# Making this call after getting the array from either of the entities (old/new) will throw the error ERROR: Max recursion reached at: write (core/variant/variant_parser.cpp:1942)
assert_that(old_entities).is_equal(backend._entities)

Throws millions of exceptions (forever). I've tried to introspect the objects in the dictionary but I can't for the life of me work out what reference if circular and what property on PandoraEntity is being introspected. Throwing this up to get feedback on the approach and to see if anyone can help debug why I can't do an is_equal comparison on them.

My hypothesis is that the value loaded into .property_map has a reference to itself but it's hard to print the stack and I can't breakpoint since there's no error, just a bunch of warnings in a loop.

Well, I'm not sure what changed between last night and this morning but apparently the tests run now without the recursion error. Haven't changed any code at all so no idea what was happening there. 🤷‍♂️

LGTM - @imothee happy for me to merge this or is there something missing?

I'm not sure if we want consistency to use _type and _value in the dictionary that's serialized (like the settings) and honestly I'm not sure if the right way to do this is to just ensure that we have access to settings and use it to set the type/hydration rather than writing it with ever single serialized value. This will at least let you have multi-typed arrays and it seems to work (been testing it at length in my game) so I think it's just a high level decision on how you foresee it working in your original architecture.

Let's merge this - in case there are any issues with this approach we can refactor it.