I like the variable approach very much, but in the example
in Markus' messageI don't understand how the reference to "$any$1" resolves to
"$any$1+age".
Would it be better to keep a clear mapping between the
variable XRIs and the XDI statements they represent? For
example:
=giovanni
+friend
$any$1
$any$2
$any$1
$context
/
+age
$greater
"40"
$any$2
$context
/
+age
$lesser
"27"
=Drummond
From:
[mailto:] On Behalf Of Markus
Sabadello
Sent: Wednesday, April 16, 2008 9:21 AM
To:
Giovanni Bartolomeo
Cc:
Subject:
Re: [xdi] $or and $and may not be necessary
Yes.. I think everyone who has some experience in logic will find
this pattern familiar.
In some cases this disjunctive normal form will
appear to be more complicated than your original form with brackets, for
example:
"Give me all of Giovanni's friends who are younger than 27 OR
older than 40". In order to implement the OR, you have to introduce two
variables, i.e. you have to do it this
way:
=giovanni
+friend
$any$1
$any$2
$any$1+age
$greater
"40"
$any$2+age
$lesser
"27"
Whereas in your
original bracket form you would probably be able to somehow do this with just
one variable.
BTW you don't need $not either, because you can convert
your $not queries like this:
"Give me all of Giovanni's friends who are NOT
(younger than 27 OR older than 40)".
Well that's just the same
as
"Give me all of Giovanni's friends who are >= 27 AND <=
40.
It really seems to be just a simple matter of boolean algebra to
convert just about any arbitrarily complex query to what you call "normal
form".
Markus
P.S. You can try the query above at http://graceland.parityinc.net/xdi-querier/XDIQuerier
On Wed, Apr 16, 2008 at 1:47 PM, Giovanni Bartolomeo
<>
wrote:
Coming
back a bit to this thread... I've made some thoughts.
The solution
proposed by Markus works because he has been able to express this complex
query into a "normal form" compliant with the one used for "traditional"
$get. He correctly identified that this format allows to disjoin the graph
into multiple graphs (he called them "clusters"). Adding different $any
variables, we can implicitly create AND and OR statements, in a simple way:
AND statements are created when the same $any variable appears into two or
more clusters, OR statements are the default.
So this solution will work
in general for any AND/OR expression. The only concern is that we've to
arrange the query into this "normal form". Interestingly, the existence of a
"normal form" for XDI has reminded me some expressions in logic which are
known as disjunctive normal forms: http://en.wikipedia.org/wiki/Disjunctive_normal_form...
not sure, but maybe it worths to be further investigated in the
future.
Bests,
Giovanni
At 15.40 01/04/2008, Giovanni Bartolomeo
wrote:
Hello Markus,
at a first glance it seems that
this solution works fine, however I'm
not sure whether this could
replace $and, $or etc. in any possible use
case (we need to elaborate
some more examples).
But I agree that for now this is a good solution
and we can reccomend
to use it :-)
Giovanni
At 20.46
30/03/2008, you wrote:
Maybe $or and $and are not necessary at all,
because I think that
1) There's an implicit AND relation between query
statements that
share variables.
2) There's an implicit OR relation
between query statements that do
not share variables.
For
example, take this query:
=giovanni
+friend
$any$1
+brother
$any$2
$any$1+age
$lesser
"40"
It means give me all statements that are matched by the
following:
( [=giovanni[+friend[$any$1]]] AND
[$any$1+age[$lesser["40"]]] ) OR (
[=giovanni[+brother[$any$2]]]
)
In other words, you can divide the query graph into multiple
disjunct
graphs (let's call them clusters). Every variable you use
appears in
only one cluster. The query above consists of two
clusters:
Cluster 1:
=giovanni
+friend
$any$1
$any$1+age
$lesser
"40"
Cluster
2:
=giovanni
+brother
$any$2
Now the rule is that when executing the query you
select everything
from your input graph that matches ALL statements in
AT LEAST ONE of
the clusters. This results in the implicit AND and OR
relations.
I'm not 100% sure, but I think that with Giovanni's $any
and $all
words you can automatically model any query you want, without
the need
for $or and $and.
Markus
P.S. I'm sure that in
graph theory there's a better word for those
"clusters", but I don't
know the correct expression in either English
or German. The point is
that variables that are not "connected" end up
in different
clusters.
---------------------------------------------------------------------
To
unsubscribe from this mail list, you must leave the OASIS TC
that
generates this mail. You may a link to this group and all
your TCs in OASIS
at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
---------------------------------------------------------------------
To
unsubscribe from this mail list, you must leave the OASIS TC
that
generates this mail. You may a link to this group and all your
TCs in OASIS
at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
