Re: Proposal for matching negated sets (was Re: New Public Review Issue: Proposed Update UTS #18)

From: Hans Aberg (
Date: Tue Oct 09 2007 - 12:09:29 CDT

  • Next message: Mike: "Re: Proposal for additional syntax (was Re: New Public Review Issue: Proposed Update UTS #18)"

    On 9 Oct 2007, at 17:06, Philippe Verdy wrote:

    >>> So the question becomes really: What does a negated *regexp* match?
    >> The link I gave suggested: the complement language.
    > Apparently you did not follow the discussions. This is just the
    > intuitive
    > answer, but not enough to determine the effective behaviour!
    > The "complement" language can only be understood if you know what
    > is the
    > "complete" language, ...

    In the link I gave, the vocabulary V is the character set.

    > ...which is directly linked to the way texts are modelized
    > and matched (including for special cases like aggregates: consider
    > the case
    > of "a*" or "a+", and think about what can be their complement,
    > because it is
    > exactly the same problem, if no semantic is defined for the "complete"
    > language).

    And in the link I gave, if you know the associated DFA, the
    complement DFA can be constructed, which is then matches the
    complement language in V*.

    > You can't answer to these questions as it interacts with the possible
    > (optional and unspecified) rules of the longest match or leftmost
    > match if
    > they are implemented, or with the ordering of matches when multiple
    > matches
    > are returned in a sequence or with enumerating methods like
    > "regexp.nextMatch()" or "regexp.previousMatch()".

    What is you point? - DFAs can be constructed for those.

       Hans Åberg

    This archive was generated by hypermail 2.1.5 : Tue Oct 09 2007 - 12:12:56 CDT