Re: RM and abstract syntax trees
Date: Thu, 01 Nov 2007 11:37:08 GMT
Message-ID: <ETiWi.2623$hd1.1727_at_trndny01>
"Marshall" <marshall.spight_at_gmail.com> wrote in message
news:1193888146.598668.286650_at_v29g2000prd.googlegroups.com...
> Actually, Paul had a pretty good take on atomic just now:
> it means the relational operators can't decompose them.
> As has been said a bunch of times, the only operation
> the RM requires on attribute domains is the equality test.
We've had this conversation before. But here goes.
The equality test is most often implemented on "simple" data types by testing on equality of the representations, not on equality of the values thus represented. This works, provided the representation scheme does not allow synonyms, i.e. two distinct representations for the same thing. If synonyms are present, the equality test on representations can yield false negatives, if what is really desired is an equality test on values.
Most of the representation schemes for compound values, like RVAs, permit synonyms. The ones that don't permit synonyms are more costly. Received on Thu Nov 01 2007 - 12:37:08 CET