IsamU:
Common theoretical computer science tradition seems to treat RE
concatination as a logically associative operator, then the
grammar may make it either left-associative or
right-associative.
This is a good argument for what was (as I recall off the top of my
head) interpretation (2) in the RE-ASSOC question. Specifically,
associativity is another desirable algebraic property which Posix
regexps might conceivably have, but under both variations on (1), they
don't have that property.
I think the larger problem with (2) is that it presumes too much that
just isn't supported by other language in the spec. It would be
something new, and interesting, but not Posix regexps.
I'll leave it to someone else to chase down the references Paul gave
and see what they say about regexp languages which aren't associative.
Incidentally, I like to promote the use of more precise terminology in
this area:
regular expressions: the simplest class in the Chompsy
hierarcy; the languages Kleene studied.
regexps: the more complicated pattern languages used in, for
example, Perl and Posix.
-t
|