This week I try to atone for my sins.

Proper Properties

I’ve got to get better at this socializing or social media or self-presentation and promotion, right? Last weekend I transcribed my thoughts into a YouTube slideshow …

LINK >

… and I missed an issue that meant I needed to replace all of it. Quite embarrassing.

The Issue

I’m binding and “instrumenting” C++/17 classes via Template Meta Programming. I want to bind/instrument C++ code with one set of defs, to three systems, with “negligible” overhead. To do this - I’m working with method pointers. To test this, I tried to write classes in #define macro() blocks that used if constexpr to branch and return void or a pal:etry<void> error monad.

That didn’t work;

• the void methods can’t have a …

  #define RES(KEY, T)                           \
    struct KEY ## _class                        \
    {                                           \
      int _ ## KEY;                             \
      T KEY(int KEY)                            \
      {                                         \
        _ ## KEY = KEY;                         \
        if constexpr(std::is_same_v<void, T>)   \
        {                                       \
          return;                               \
        } else {                                \
          return pal:etry<void>::pass();        \
        }                                       \
      }                                         \
    };                                          \
  ...
  RES(foo1, void);
  RES(foo2, pal::etry<void>);
  

  … since C++ is too dumb / smurt enuff to notice those return types don’t match.

  Not a concern for IRL usage; just for generating test cases with macros and TMP.

• One also can’t do …

  template<class R>
  struct SUPER<R>
  {
    R bar(float bar)
    {
      ...
      if constexpr(std::is_same_v<void, R>)
      {
        return;
      } else {
        return pal:etry<void>::pass();
      }
    }
  };
  struct CHILD : SUPER<void>
  {
  };
  

  … since CHILD::bar doesn’t exist - only SUPER<void>::bar exists.

Fine.

The Fix

#define PROPERTY_TEST_CLASS(KEY, R, P)        \
  struct property_example_ ## KEY             \
  {                                           \
    int _value;                               \
    R value(int value)                        \
    {                                         \
      _value = value;                         \
      P;                                      \
    }                                         \
  };                                          \
  p_test_case(property_test_ ## KEY)          \
  {                                           \
     ...                                      \
  }                                           \

PROPERTY_TEST_CLASS(01, void,             {});
PROPERTY_TEST_CLASS(02, pal::etry<void>,  { return pal:etry<void>::pass(); });

Reflection and the Future

This is for a game engine. I (like many others) am attempting to produce a game engine - almost from scratch. Part of this involves scripting serialisation and inspection or SSI of data, or, code objects.

That’s not something I see a lot of good/usable COTS/FOSS tools for. So … a valid area of special interest?

Anyway

I put the above hijinks into a file, then, use a Python script to generate permutations of the classes and tests. It’s up to 64 variations - I also test some cases by hand.

Current Headache

The getter and setter will need to take “context” data. I’ve been reluctant to “lock in” to one datatype … but as of this writing …

It should just be an optional entt::registry& parameter. All signatures should/can/do take an optional entt::registry& parameter.

• Duktape callbacks can live with this; they can retrieve it
• Duktape might be retired/replaced with Lua
• entt has a conventional entt::registry::ctx<T>(): T&

… again; F I I N N N E E E E

Concludes

So that’s my weekend planned;

• Resident Evil / Endgame (IDK)
• SprueCon
• set-up hgkeeper
  • … and backups …
  • … Drone CI …
  • … and mirror to … idk … sr.ht? HgLab?
  • … and some RO Kanban Board
• sweep-through my project tree
  • pal::context is RO and init from `pal::context