Re: Automata builder -- interface

From: dg_at_[hidden]
Date: 2001-05-31 04:04:08

I'm a bit of an outsider to this group, but I'd like to venture a
comment here anyway.

Vladimir Prus wrote:
 > Another possible purpose of an automaton is to perform transitions
 > from state to state (and possibly output something in the
 > process). Actually, the word "primary" that I have used should be
 > dropped -- it's reasonable to have both Acceptor interface and
 > transition interface available to the world.

Why not also view the acceptor case as having an output of type
bool. Then every automaton could have a template parameter to specify
it's output type.

But just specifying an output type would be inadequate. You also need
to know how the output types get combined. If the output type is
string, it might be reasonable to presume that these outputs would be
concatenated. But what if the output type is float or double?

To get some idea as to how to handle various kinds of outputs, I would 
suggest that you look at the nice automata classes developed at AT&T:

   http://www.research.att.com/sw/tools/fsm/

-- Dale Gerdemann