On Thu, Nov 29, 2012 at 10:12 AM, Graydon <graydon@xxxxxxxxx> wrote:
> On Thu, Nov 29, 2012 at 05:58:45AM +0000, Ihe Onwuka scripsit:
>> > Since we're talking about "except" and language definitions, has anyone
>> > ever proposed a different form (or is there a different form for) the
>> > construct:
>> >
>> > div[not(* except (heading,para))]
>> >
>> > It's really useful, but I find it frequently brain-melting, especially
>> > as a component of more complex expressions.
>>
>> A problem here is the mixture of notation from mathematical (set
>> theoretic in the form of except) and non-mathematical (the comma
>> operator).
> [snip]
>> Now (get ready for the leap of faith and giant hand wave), substitute
>> the comma operator back for the | and you have div[(heading,para]
>
> All the divs who have heading or para children, yes, but that's not
> quite the same thing.
>
> Unless I'm unusually addled this morning,
>
> div[not(* except (heading,para))]
>
> is a div which has _only_ heading or para children
>

Well the problem is with the semantic effect of div [ ] not the set
theoretical translation of the predicate.

Returning to my earlier notation  take the complement of what we ended up with

 (heading|para)'

Only - would be things that are not in the above expression - we are
in to double negative territory.

div[not(heading|para)']   (that should bother you - well it does me)

but you could do a mechanical DeMorgan translation to turn that into
something that keeps the computer happy

div[(not ((* except heading ) union (* except para))]

 while you mentally continue to parse it as  div[not(heading|para)']