import std.array; void main() { void[] a = cast(void[]) [ 1, 2, 3 ]; a.front; } yields: ..\src\phobos\std\array.d(419): Error: [i] has no effect in expression (a[0u]) ..\src\phobos\std\array.d(5): Error: template instance std.array.front!(void[]) error instantiating Originally I used std.file.read(), hence the void array. - 1st message is completely senseless, what "[i]"? - 2nd message shows line 5 which is the line of "a.front;" but the filename is wrong.
This is actually two separate issues. The first is a Phobos issue, namely that there should be a template constraint on std.array.front() that prevents it from being instantiated on void[] arrays. front() is defined as something like T front(T)(T[] array) { return array[0]; } You cannot index into a void[] array, hence the first error message. This also means that void[] cannot be a range. The second is the issue of the wrong line number, which is a bug in DMD. I'm guessing it's just bug 1913 again, so I'm marking this one as a Phobos issue. In the case of a void[] returned from std.file.read(), you need to cast it to ubyte[] before using it as a range.
This fixes the Phobos bug: http://www.dsource.org/projects/phobos/changeset/2123 The error message is now: test.d(6): Error: template std.array.front(A) if (is(typeof(A[0])) && !isNarrowString!(A) && !is(typeof(A[0]) : const(void))) does not match any function template declaration test.d(6): Error: template std.array.front(A) if (is(typeof(A[0])) && !isNarrowString!(A) && !is(typeof(A[0]) : const(void))) cannot deduce template function from argument types !()(void[]) Arguably less readable, but more useful since it refers to the point of error in user code instead of in Phobos code.
(In reply to comment #2) > Arguably less readable, but more useful since it refers to the point of error > in user code instead of in Phobos code. Too bad there is no handy way to add error messages to template constraints...
Thx for the fix, but I still think that strange [i] in the error message is a diagnostic bug, reopen it?
I think '[i]' is just a compact way of saying 'indexing operation'. I guess it could just as well have been '[#]', '[index]', or whatever. (Even better would be a message saying that you can't index into void[] arrays.) Anyway, I suggest you open a new bug report about it. It's better not to have two bugs in one report -- especially when one is a Phobos bug and the other is a DMD bug.