Re: If you were to implement the original relation algebra language...

Joe \"Nuke Me Xemu\" Foster wrote:
> "Amund Trovåg" <amundAndHereEndethMyName_at_removeINOLIKESPAMYOUSEEtexassibir.com> wrote in message
> <news:3FBD12DB.40205_at_removeINOLIKESPAMYOUSEEtexassibir.com>...
>
>

>>Would precedence have a lot to say for optimization?
>>
>>I am currently working on doing just this, creating the grammar file
>>from BNF to JavaCC(a program that makes Java parser files for me, based
>>on a grammar).
>>
>>I am not certain whether project should have precedence over the other
>>operators, as this might create trouble when combined with e.g. selection.
>>
>>Any thoughts, comments or tips on how the grammar should be structured
>>with regard to this?

> It shouldn't impede optimization, and
> a functional notation for those of us with "standard" keyboards will
> all but demand it:
>
> URL:http://www.cs.rochester.edu/users/faculty/nelson/courses/csc_173/relations/algebra.html
>
> You might be able to get away with R1 + R2 for union, R1 - R2 for
> set difference, R1 * R2 for kernel panic: out of swap, etc., but you
> should encourage parentheses here too, because R1 - (R1 - R2) isn't
> the same as (R1 - R1) - R2.

Yeah you got a point there. Association must be looked at when deciding for parentheses.
I am actually going to allow the user to say "project" "union" etc, instead of using symbols... Of course this might be a bit of a trick if we have an attribute called e.g. "union", but in my environment it wont be a problem though.

Amund Received on Thu Nov 20 2003 - 21:20:51 CET