• cc (27/27) Aug 25 2014 Is it possible to allow implicit casting from a base type to a
• Jonathan M Davis (4/6) Aug 25 2014 All implict conversions are done using alias this:
• "Marc =?UTF-8?B?U2Now7x0eiI=?= <schuetzm gmx.net> (5/9) Aug 26 2014 There is currently no implicit for aggregates except using `alias
• cc (20/20) Aug 26 2014 Ahh, thanks. Looks like encapsulating the union in a struct with

"cc" <cc nevernet.com> writes:

Is it possible to allow implicit casting from a base type to a
union type?  For example, considering the following basic vector
union:

union vec2 {
	struct {
		float x = 0.0f;
		float y = 0.0f;
	}
	float[2] v;
	enum length = v.length;

	ref auto opIndex(size_t idx) {
		assert(idx >= 0 && idx < length, "Bounds error on index");
		return v[idx];
	}
	auto opAssign(float[] f) {
		assert(f.length == length, "Bounds error on assignment");
		v[0..length] = f[0..length];
		return this;
	}
}

vec2 a = vec2(1.0, 2.0); // fine
vec2 b;
b = [3.0, 4.0]; //fine
vec2 c = [5.0, 6.0]; // cannot cast float[] to vec2


Overloading opCast() seems to be only for "outbound" casting, and
unions can't have this() constructors like structs.  Is there any
way to accomplish this?

"Jonathan M Davis" <jmdavisProg gmx.com> writes:

On Tuesday, 26 August 2014 at 02:33:25 UTC, cc wrote:
 Is it possible to allow implicit casting from a base type to a
 union type?

All implict conversions are done using alias this:

http://dlang.org/class.html#AliasThis

- Jonathan M Davis

"Marc =?UTF-8?B?U2Now7x0eiI=?= <schuetzm gmx.net> writes:

On Tuesday, 26 August 2014 at 02:33:25 UTC, cc wrote:
 vec2 a = vec2(1.0, 2.0); // fine
 vec2 b;
 b = [3.0, 4.0]; //fine
 vec2 c = [5.0, 6.0]; // cannot cast float[] to vec2

There is currently no implicit for aggregates except using `alias 
this`. But in your example, it's a construction, not an 
assignment, so you need to write an appropriate constructor 
`this(float[])` analogous to `opAssign(float[])`.

"cc" <cc nevernet.com> writes:

Ahh, thanks.  Looks like encapsulating the union in a struct with 
alias this gets the job done, and removes the need for overloads. 
  Neat.

struct vec2 {
	union {
		struct {
			float x = 0.0f;
			float y = 0.0f;
		}
		float[2] v;
	}

	this(float x, float y) {
		this.x = x;
		this.y = y;
	}
	this(float[v.length] f) {
		v[0..length] = f[0..length];
	}
	alias v this;
}