Re: FYI: Regex paper for UTC

From: Hans Aberg (
Date: Sun Oct 14 2007 - 02:18:07 CDT

  • Next message: Hans Aberg: "Re: FYI: Regex paper for UTC"

    On 14 Oct 2007, at 00:51, Philippe Verdy wrote:

    > I don't know why you want to restrict the negation to only a small
    > part of
    > what they can do.

    You can define other operations that the ordinary complement; just
    don't all cpmplement, lest peope will become confused.

    > My negation operator (like all other regexp operations) IS
    > operating as a
    > injection from a set of positioned strings=(C*, P) to (C*, P) where
    > P is a
    > set of positions within strings of C*. It is more general and
    > solves the
    > particular case of injections from C* to C* (where P is then only a
    > singleton based on a particular rule).

    You make it complicated: Once the language L(x) has been defined,
    just find the longest match in the string that is in the language.

    > Note: my engine does not look first for the longest match.

    So the there should some oether match in the language.

    > Longest matches
    > are found by filtering, using customizable rules after each matche
    > (whatever
    > its size or position) is found. Why?:
    > Consider an input text coming from a stream A that generates an
    > infinite
    > suite of characters "a".
    > Feed this stream to a filter that will return the longest match for
    > the
    > regexp "a*". This filter will never return unless it applies the
    > technic of
    > optimization for the conversion of a final loop into a final
    > recursion.
    > Now use the same filter but with the regexp "a*b": this time it's
    > impossible
    > for the filter to return anything.

    But the normal thing is that one wants it to eat it all. And if one
    wants partial returns, that can be done with a lazy implementation.

    It seems me that you are not merely wanting to extend regular
    expression, but want to tie that to a particular implementation: your

       Hans Åberg

    This archive was generated by hypermail 2.1.5 : Sun Oct 14 2007 - 02:19:50 CDT