No equivalent "IsDestroyed" method in World interface

[opera-aberglund]

The "IsDestroyed" methods in AabbTreeWorld doesn't exist in the World interface. For example:

bool IsDestroyed(const AabbTreeWorld& world, BodyID id) noexcept
{
    return world.m_bodyBuffer.FindFree(to_underlying(id));
}

There should be something like:

bool IsDestroyed(const World& world, BodyID id)
{
    return world.m_impl->IsDestroyed_(id);
}

and equivalent in the WorldConcept and WorldModel...

[louis-langholtz]

Out of curiosity, what flavor of interface for this info would you prefer:

1. IsDestroyed(const World&, EntityID).
2. entity.IsDestroyed().
3. std::optional<Entity> GetEntity(const World&, EntityID).

[opera-aberglund]

Out of curiosity, what flavor of interface for this info would you prefer:

1. IsDestroyed(const World&, EntityID)
2. entity.IsDestroyed()
3. std::optional GetEntity(const World&, EntityID)

I would prefer 1 I think because it matches the signature of the Destroy function.

[louis-langholtz]

I would prefer 1 I think because it matches the signature of the Destroy function.

Interesting to me that option 3 didn't get mentioned. Interesting in terms of discussion, because while I'm a fan of C++ vocabulary types like std::optional, I'm also concerned about minimizing memory usage.

Specifically, making the underlying storage arrays that the implementing world uses be arrays of std::optional<T> instead of arrays of T seems more likely to be suboptimal bit-density wise than another possibility I'm thinking of.

In this case, PlayRho types like Body and Contact already provide boolean properties stored directly in bits packed optimally. That leads most naturally for me to options 1 and 2.

The more abstracted PlayRho types Shape and Joint don't seem to have as optimal, or, at least, not such a straight forward path for providing an is-destroyed piece of info. Perhaps the most straight forward to me is using their has_value member function for this. That way, supporting types (like DiskShapeConf for example), don't have to all change their interfaces to support this.

So I think that's what I'll do then, and why. I.e. provide interface 1 (and 2) for the PlayRho Body, Contact, Joint, Shape types. Assuming of course that this works out; I'm concerned I may yet encounter wrinkles with this design.

Relatedly from my thinking, and see issue #547, I'm continuing to consider removing all of the World based functions GetBodies(const World&), GetContacts(const World&), GetJoints(const World&) (and their implementation based counterparts).

[louis-langholtz]

The following is available now in current master branch:

• IsDestroyed(const Body&).
• IsDestroyed(const Contact&).
• IsDestroyed(const Shape&).
• IsDestroyed(const Joint&).
• IsDestroyed(const World&, BodyID id).
• IsDestroyed(const World&, ContactID id).

Relatedly, manifolds would use the info for contact.

I still have to add the following however - I'd forgotten these but meant to add these as well:

• IsDestroyed(const World&, JointID id).
• IsDestroyed(const World&, ShapeID id).

[opera-aberglund]

The following is available now in current master branch:

Funny thing I just realised I don't need them anymore since I just use the MakeTouchingMap function now to filter the contacts. I think it's good to have them anyway!

[louis-langholtz]

This is complete now with PR #566.