Miriad datasets are implemented as a directoryB.1. The data itself are organized in items, normally implemented as separate files, but small items (32 bytes or less) can be found together in a file called header. The Miriad program itemize will list the items in a dataset. Other programs that manipulate items are puthd (add of modify a simple item), copyhd (copy an item from one dataset to another), delhd (remove an item), gethd (show value of a simple item), prthd (show compound contents of a dataset), and mathd (perform a mathematical operation on an item).
Miriad currently knows about two types of data: visibility data and image cubes, described in a bit more detail below: