Bug 4880 – [patch] std.range.put does not handle output ranges implement via opDispatch; breaks on RefAppender

std.range.put does not correctly introspect types which implement put via opDispatch. This is fairly simple to fix by swapping hasMember out for a __traits compiles test. void put(R, E)(ref R r, E e) { + static if (__traits(compiles, {return R.init.put(E.init);}) ) - static if (hasMember!(R, "put") || - (isPointer!R && is(pointerTarget!R == struct) && - hasMember!(pointerTarget!R, "put"))) {

This is a rather problematic bug. If we use your patch, it will break when uniform function call syntax is fully implemented. I guess what we need is for hasMember to be less restrictive. It should include opDispatch, and while we're at it, pointers to structs.

Well half of puts' functionality is to emulate uniform function call syntax for structs and classes: i.e. re-write put(r,e) into r.put(e). So I'd expect UFC to break put with or without this patch. I had thought of improving hasMembers (either the __traits or templated version) to support opDispatch, however, I do not think that it is possible. The problem is template constraints. Consider the highly synthetic example: void opDispatch(string name, T...) if( name.length == T.length ) {} Because the template constraint could be non-trivially dependent on something more than name there's no way to check if r.opDispatch!"put" is valid without actually evaluating r.put(e) (i.e. __traits(compiles,...))

I converted the patch into a pull request: https://github.com/D-Programming-Language/phobos/pull/433

I close this, because the bug is very old, the code (for ranges) has had lots of changes since then and no example is provided. I also was not able to provide an example myself, but I cannot say, if this was due to my missing skills or the bug being fixed meanwhile. If anyone can provide an example, please feel free to reopen!