I've added some Datalog material to the wiki:
http://code.google.com/p/clojure-contrib/wiki/DatalogOverview
On Wed, Feb 18, 2009 at 4:55 PM, Jeffrey Straszheim <
straszheimjeff...@gmail.com> wrote:
> Makes sense. That would work. It certainly looks cleaner.
>
>
> On Wed, Feb 18, 2009 at 4:51 PM, Rich Hickey <richhic...@gmail.com> wrote:
>
>>
>>
>>
>> On Feb 18, 4:32 pm, Jeffrey Straszheim <straszheimjeff...@gmail.com>
>> wrote:
>> > Easy enough to do. The only drawback is I'd probably want to force it
>> into
>> > a hash during the query. For large datasets (say 100,000 records) this
>> > might get expensive.
>> >
>>
>> What I envisioned was that while this was the logical db, 'inserting'
>> any tuple would also add it to internal indexes, including of course a
>> default index on relname. You'd declare which other keys to index, and
>> how (sorted/hashed) when creating the db.
>>
>> Rich
>>
>> > On Wed, Feb 18, 2009 at 4:13 PM, Rich Hickey <richhic...@gmail.com>
>> wrote:
>> >
>> > > On Feb 18, 3:51 pm, Jeffrey Straszheim <straszheimjeff...@gmail.com>
>> > > wrote:
>> > > > Yes. I've been thinking about a database layer that would support
>> > > indexing,
>> > > > constraints, and so on. One step at a time.
>> >
>> > > Maybe I wasn't clear, I'm talking about the foundational layer.
>> > > Instead of:
>> >
>> > > (def data {
>> > > :table-1 #{ { :x 34 :y 33 } { :x 33 :y :fred } { :x "k" :y \u } }
>> > > :table-2 #{ { :a "fred" :b "mary" } { :a "sally" :b "joan" } }
>> > > })
>> >
>> > > I'm recommending:
>> >
>> > > (def data
>> > > #{{:rel :table-1 :x 34 :y 33 }
>> > > {:rel :table-1 :x 33 :y :fred }
>> > > {:rel :table-1 :x "k" :y \u }
>> > > {:rel :table-2 :a "fred" :b "mary"}
>> > > {:rel :table-2 :a "sally" :b "joan" }})
>> >
>> > > i.e. making relation and rule names non-special.
>> >
>> > > Rich
>> >
>> > > > (logic-rule (:fred :x ?x :y ?y) - (:sally :x ?x :z ?z) ("becky" :y
>> ?y)
>> > > > (not! :janet :qqq ?z) (if < ?x
>> ?y))
>> >
>> > > > Translated into positional notation (assuming the columns are named
>> in
>> > > the
>> > > > obvious way):
>> >
>> > > > fred(X,Y) :- sally(X,Z), becky(Y), ~Janet(Z), when X<Y.
>> >
>> > > > The "<" symbol can be any Clojure callable. Its return value will
>> be
>> > > > interpreted as a boolean.
>> >
>> > > > So, you'd get every X,Z from the relation sally, cross product with
>> every
>> > > Y
>> > > > from becky, remove each tuple that has a Z in janet, and also remove
>> any
>> > > > tuple where X<Y fails. The resulting X,Y would be projected as
>> relation
>> > > > fred.
>> >
>> > > > On Wed, Feb 18, 2009 at 3:23 PM, Rich Hickey <richhic...@gmail.com>
>> > > wrote:
>> >
>> > > > > On Feb 9, 8:46 am, Jeffrey Straszheim <
>> straszheimjeff...@gmail.com>
>> > > > > wrote:
>> > > > > > No, but I'm really learning as I go here. I'll look into it.
>> >
>> > > > > > On Mon, Feb 9, 2009 at 7:58 AM, Rich Hickey <
>> richhic...@gmail.com>
>> > > > > wrote:
>> >
>> > > > > > > Looks like you're moving apace!
>> >
>> > > > > > > Have you considered query/subquery optimization instead of
>> magic
>> > > sets?
>> >
>> > > > > > > Rich
>> >
>> > > > > > > On Feb 8, 7:51 pm, Jeffrey Straszheim <
>> straszheimjeff...@gmail.com
>> >
>> > > > > > > wrote:
>> > > > > > > > By the way, if anyone on this list has experience
>> implementing
>> > > > > bottom-up
>> > > > > > > > optimizations for logic programs, particularly from the
>> magic set
>> > > > > family,
>> > > > > > > > and is willing to assist, please contact me.
>> >
>> > > > > > > > On Sun, Feb 8, 2009 at 7:47 PM, Jeffrey Straszheim <
>> >
>> > > > > > > > straszheimjeff...@gmail.com> wrote:
>> >
>> > > > > > > > > Stratified negation is working and in trunk.
>> >
>> > > > > > > > > I have some cool ideas of a simple, but powerful, way to
>> > > implement
>> > > > > > > > > evaluable predicates. They'll likely make it in by
>> midweek.
>> >
>> > > > > > > > > The the hard part (magic sets) begins.
>> >
>> > > > > > > > > On Feb 8, 11:43 am, Jeffrey Straszheim <
>> > > > > straszheimjeff...@gmail.com>
>> > > > > > > > > wrote:
>> > > > > > > > > > I now have recursive queries working. My next 3
>> milestones
>> > > are
>> > > > > > > > > stratified
>> > > > > > > > > > negation, evaluable predicates, and then some version of
>> > > magic
>> > > > > sets
>> > > > > > > > > > optimization. But now, as long as your queries are
>> > > non-negated
>> > > > > it is
>> > > > > > > > > > working.
>> >
>> > > > > > > > > >http://code.google.com/p/clojure-datalog/
>> >
>> > > > > I got a chance to look at your docs:
>> >
>> > > > >http://code.google.com/p/clojure-datalog/wiki/BasicSyntax
>> >
>> > > > > I think your choice of using maps (we don't call them hashes in
>> > > > > Clojure as they might not be hash tables) is right on the money
>> for
>> > > > > Clojure, especially set-of-maps-is-relation, just like
>> clojure.set.
>> >
>> > > > > Two thoughts:
>> >
>> > > > > I wonder though if the map of rel-names to rels isn't a wart
>> though.
>> > > > > It's a pet peeve of mine that relation names don't end up in the
>> db
>> > > > > like any other attribute. Yes, they'll need to be indexed, but
>> > > > > eventually you'll want to support indexing on any desired
>> attributes
>> > > > > as well. Putting relation names in the db gives you a uniform
>> meta-
>> > > > > query capability. I haven't thought this all the way through, but
>> you
>> > > > > might want to think about it.
>> >
>> > > > > I didn't know how to interpret this:
>> >
>> > > > > (logic-rule (:fred :x ?x :y ?y) - (:sally :x ?x :z ?z) ("becky" :y
>> ?y)
>> > > > > (not! :janet :qqq ?z) (if < ?x
>> ?
>> > > > > y))
>> >
>> > > > > Overall, it looks very promising!
>> >
>> > > > > If you weren't aware of it:
>> >
>> > > > > Foundations of Databases: The Logical Level
>> > > > > Serge Abiteboul, Richard Hull, Victor Vianu
>> >
>> > > > >http://www.amazon.com/gp/product/0201537710
>> >
>> > > > > has good coverage of Datalog, including QSQ.
>> >
>> > > > > Rich
>> >>
>>
>
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Clojure" group.
To post to this group, send email to clojure@googlegroups.com
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
-~----------~----~----~----~------~----~------~--~---
