[sc34wg3] Fwd: Re: TMQL: Navigation "traverse"

Lars Heuer heuer at semagia.com
Tue Nov 4 13:17:14 EST 2008


This is a forwarded message

From: Robert Barta <rho at devc.at>
To: Lars Heuer <heuer at semagia.com>
Date: Tuesday, November 4, 2008 7:03:29 PM
Subject: [sc34wg3] TMQL: Navigation "traverse"

===8<==============Original message text===============

On Tue, Nov 04, 2008 at 05:53:12PM +0100, Lars Heuer wrote:
> Hi Robert,
> 
> [Removal of player topic once in the "traverse" axis]
> > But this is certainly something one should think about one time more.
> 
> What was the initial reason to add this rule? For binary associations,
> that only the other players are in the result? If this was the reason,
> it wouldn't work well if the player is involved in more than one
> association of the same type.

No, it *would* work, as the 'once' is to be interpreted (phrasing
not perfect) _per traversed association.

> I am irritated by the "once" a bit, either the player should be
> removed from the result (only the other players are part of the
> result) ...

I did not like that because it would distort the results in
an non-monotonous way:

  publication (author: lheuer, editor: lheuer, author: rho)

lheuer <-> publication          rho

rho    <-> publication          lheuer, lheuer

  publication (author: lheuer, editor: xxx, author: rho)

lheuer <-> publication          rho, xxx

rho    <-> publication          lheuer, xxx

Only the content changed (but not any cardinality), and still the
cardinality of the query results differ.

And "remove the incoming player" it would introduce a 'set' semantics.
This would be the one and only place where that would happen in TMQL.
Very unelegant. And it would make a larger(!) mess when writing the
formal semantics for it.

> or the result should be unmodified (the player is not removed
> from the result).

More orthodox but highly impractical. It would question the existence
of <->, IMHO.

\rho




===8<===========End of original message text===========



More information about the sc34wg3 mailing list