Typed GML

Typed GML is an attempt to add typed functions, constructors and methods in the new Game Maker Studio 2.3 beta.

Supported type checks:

• check if a struct came from a specific constructor (t_constructor_of)
• instance object_index checks (t_instance_of)
• all built in types functions (real, array, string, numeric, method etc)

When and why to use?

Type checking gets rid of the worst bugs by stopping any buggy code right where the bug occurs. For example if you made a typo in your code some days ago and have been working on unrelated things, but then find an undefined variable showing up in a random place, there is no simple way to know what caused it. Many functions handle undefined values just fine (e.g. a list, grid set instead of a multiplication) and the source of a bug can hence be hard to track down.

The aim of Typed GML is to solve that by letting the programmer assert input and output types. This is especially useful when dealing with unknown data, large data pipelines, networking or any other more complex application. For release mode, type assertions can be turned off by modifying a global macro.

Example Usage

globalvar Entity;
Entity = function(_name, _instance) constructor {
	assert_types(t_string(_name));
	
	name = _name;
	instance = _instance;
}

function SimpleStorer(_instance) constructor {
	assert_types(t_instance_of(_instance, obj_simple));
	
	instance = _instance
}

function Vec2(_x, _y) constructor {
	assert_types(t_real(_x), t_real(_y));
	
	x = _x;
	y = _y;
	
	static Add = function( _other ){
		assert_types(t_struct_of(_other, self));
		
		x += _other.x;
		y += _other.y;
	}
}

globalvar Struct;
Struct = function(_entity, _vector) constructor {
	assert_types(t_struct_of(_entity, Entity), t_struct_of(_vector, Vec2));
	
	entity = _entity;
	vector = _vector;
	
	static get_entity = function() {
		return entity;
	}
	
	static set_entity = function(_entity) {
		assert_types(t_struct_of(_entity, Entity));
		entity = _entity;
	}
}

Performance

The project has performance and unit tests. These were run on a mac and performance may differ between platforms. If someone has another platform to test this on and send the results as an issue, it would be helpful.

Type assertions can be turned off via a macro for release mode.

Mac, VM (Macbook Pro 2015)

RUNNING PERFORMANCE TESTS 
TEST: assert a real variable , running 25000 times 
0 took 87.29 microseconds per one run, ops per step (60fps): 191 
1 took 101.55 microseconds per one run, ops per step (60fps): 164 
extra time from type assertions:  14.26 microseconds per one run, ops per step (60fps): 1169 

TEST: assert 4 real type variables , running 25000 times 
0 took 76.19 microseconds per one run, ops per step (60fps): 219 
1 took 102.55 microseconds per one run, ops per step (60fps): 163 
extra time from type assertions:  26.36 microseconds per one run, ops per step (60fps): 632 

TEST: assert 16 real type variables , running 25000 times 
0 took 67.67 microseconds per one run, ops per step (60fps): 246 
1 took 127.75 microseconds per one run, ops per step (60fps): 130 
extra time from type assertions:  60.08 microseconds per one run, ops per step (60fps): 277 

TEST: assert a vec2 struct type , running 25000 times 
0 took 64.13 microseconds per one run, ops per step (60fps): 260 
1 took 78.78 microseconds per one run, ops per step (60fps): 212 
extra time from type assertions:  14.65 microseconds per one run, ops per step (60fps): 1138 

TEST: for comparison, a ds_grid lookup , running 25000 times 
0 took 0.62 microseconds per one run, ops per step (60fps): 26918 

TEST: for comparison, a ds_map_create and destroy , running 25000 times 
0 took 4.13 microseconds per one run, ops per step (60fps): 4039 

TEST: for comparison, an instance_create and destroy , running 25000 times 
0 took 46.02 microseconds per one run, ops per step (60fps): 362 

TEST: for comparison, a ds_map lookup , running 25000 times 
0 took 0.80 microseconds per one run, ops per step (60fps): 20787 

TEST: for comparison, a ds_grid write , running 25000 times 
0 took 0.15 microseconds per one run, ops per step (60fps): 112067 

TEST: for comparison, a ds_map write , running 25000 times 
0 took 0.79 microseconds per one run, ops per step (60fps): 20989 

TEST: for comparison, an anonymous function call , running 25000 times 
0 took 0.89 microseconds per one run, ops per step (60fps): 18663