Cart
[Home] [By Thread] [By Date] [Recent Entries]
From: "Mark Endicott" <MEndicott@t...> > It appears to me that (A,B)*, A? isn't impossible. Wouldn't this lead to the > following list of possible child elements? > > (no children) > A > A, B > A, B, A > A, B, A, B > A, B, A, B, A > ... The ambiguity comes from a restriction which lets us use simple state machines to implement parsing a content model: we read one element and decide which state to go to next. If the grammar at the top is implemented as a state machine, the first transition has two A choices: this is ambiguity. There are various things we could do: we could take the lexically first choice in content model; we could rewrite the ambiguous content model into an unambiguous one (often not difficult); we could allow a lookahead to the possible states after the current As and then lookahead in the document to the element after the current one; we could implement a weaker validation which allows (a, b?)*. Or we could give up the ambiguity restruction and implement a different kind of validator: for example instead of a state machine, we could use a Petri net and follow multiple paths. Cheers Rick Jelliffe
|
