Hi,

In phobos there are places with void[] and places with ubyte[], and I can't really understand why there is this difference. I don't even see why D has a void array or void pointer anyway.
For example, in std.base64, ubyte[] is used as the basic quantity for encoding, while in std.zlib, it's void[]. Both libraries have a similar pair of functions like encode/decode or compress/uncompress. Both act on raw memory. So why not use ubyte[] all the time? Isn't any stretch of bare data on the heap in the end an array of bytes? Void[] seems like a C-anachronism to me. I have no experience with void pointers or arrays, so I am probably missing something.

Thanks,

Stephan

On 2/5/13, Stephan <stephan_schiffels@mac.com> wrote:
> In phobos there are places with void[] and places with ubyte[]

One benefit of void[] over ubyte[] is implicit conversion. E.g.:

void input(void[] arr) { }

void main()
{
    input([1, 2]);
    input([1.0, 2.0]);
    input(["foo", "bar"]);
}

You don't get that conversion with byte arrays.

And the benefit of byte[] over void[] is that (at least theoretically IIRC) the garbage collector won't scan byte arrays but it will scan void arrays. This means the GC could interpret an element in a void[] as a pointer and end up never releasing some bit of memory.

Someone should expand on this better than me, but I think that's the gist of it.