Idea spawned by this discussion. The idea is that just as writing v[b], v[!b] where b is a logical indexing vector partitions v into two disjoint sets of elements, writing v[x], x[!x] should also partition v into two disjoint sets of elements when x is a single index or a range of indices. This requires some new NegatedIndex type, together with ! methods for integers and ranges and the appropriate method for indexing arrays with NegatedIndex objects.