Re: Non-deterministic content model

[Eric Bohlman <ebohlman@earthlink.net>]

6/13/01 6:05:25 PM, Marcus Carr <mrc@allette.com.au> wrote:
>Jason Chalecki wrote:
>
>> ((a, b)*, a) essentially represents the language where the first and
>> last letter is a. Additionally, if there are b's, a and b must
>> alternate. So (a, (b, a)*) matches the same language and is
>> deterministic, correct?
>
>Yes, but the original model was allowed to finish with either a or b - the
>final a had a question mark. To model that with your approach would result
>in (a, (b, a?)*), which allows a string of b's with no intervening a's.
>Either way, you've changed the model.

Applying the subset construction to the NFA associated with the original model ((a, b)*, a?), I end 
up with a two-state DFA in which both states are final:

	A	B
0	1	-
1	-	0

but I can't figure out how to express it as a content model (or regular expression).