Re: RM and abstract syntax trees

Bob Badour wrote:
> paul c wrote:
>

>...
>> Here's my favourite nested relation, although I admit it's probably 
>> useless in practice.  It's a recursive one.  Sorry I don't have much 
>> mastery of conventional syntax, what I mean here is something like R: 
>> <attribute list> where <attribute list> is a set of attribute name, 
>> attribute type pairs and typeof is swiped from C-language:
>>
>> R: (A typeof R)
>>
>> I don't know how to display a value for R but I guess it could have 
>> either no tuples or one tuple.

Thanks for the formalism suggestions, will try to absorb.

I know you've tried to explain various flavours of this to me before, but an example that prompts this for me, only with two attributes, is:

R2: {A integer, B typeof R2}

In R2, I can see that a value for B isn't necessarily empty, but if it is, that must be the end of the "descent", for I don't see how an empty recursive relation can recurse "any further" as it were. (I think this would be the case even if R2 defined several "levels" of conventional non-recursive RVA's.)

Now, I admit I'm slow (but not heavy!) but I can see only one possible value for R.A which is an empty relation of type R and two possible values for R, one where R is empty and the other where R has one tuple where R.A is empty. I don't see how a relation defined recursively can "descend" from an empty value!

So far, it looks like a peculiar kind of constraint to me. As somebody else say, go ahead and attack it, I can take it! Received on Wed Oct 31 2007 - 01:24:00 CET