I see nothing in his code or documentation for handling negation or
stratification. Also, it appears to be a top down evaluator, and I don't
see any fixed-point or other recursion handling. I *suspect* this does not
guarantee termination over arbitrary safe rules. It is not real Datalog.
On Wed
It is worth looking at.
On Wed, Feb 18, 2009 at 12:42 PM, Telman Yusupov wrote:
>
> Could this be of any help for your development? There is now a version
> of Datalog for PLT Scheme:
>
> Software:
>
> http://planet.plt-scheme.org/display.ss?package=datalog.plt&owner=jaymccarthy
>
> Documentatio
Could this be of any help for your development? There is now a version
of Datalog for PLT Scheme:
Software:
http://planet.plt-scheme.org/display.ss?package=datalog.plt&owner=jaymccarthy
Documentation:
http://planet.plt-scheme.org/package-source/jaymccarthy/datalog.plt/1/0/planet-docs/datalog/ind
I was considering extending my Datalog work with customized evaluable
predicates, but have decided against it. The safety guarantees of Datalog
are just not worth giving up. To compensate, I have (very tentative) plans
of building some sort of logic oriented bottom up computation engine --
think
Sorry, I didn't mean to suggest that you held that opinion.
In any case, I am still learning Clojure, so I think I should restrict
myself to newbie questions until I am better at it. I hope I will have time
to implement an open-world reasoner, which would help make the discussion
concrete.
On Mo
On Feb 7, 2:25 pm, John Fries wrote:
> I agree with Jeffrey that there is no reason to have just one option.
I never suggested there ought to be only one option, nor am I trying
to argue against the utility of open-world reasoners. I merely asked,
given your assertion that open-world reasoners
reposted to correct a mistake in my use of X, Y and Z:
---
I agree with Jeffrey that there is no reason to have just one option.
Sometimes you want your reasoner to find a single model; sometimes you want
it to find all models (assuming you are reasoning over a finite set).
I've appen
I agree with Jeffrey that there is no reason to have just one option.
Sometimes you want your reasoner to find a single model; sometimes you want
it to find all models (assuming you are reasoning over a finite set).
I've appended a long rant about SAT-based reasoners and open-world
semantics. I h
There is no reason to have just one option.
On Thu, Feb 5, 2009 at 3:59 PM, Rich Hickey wrote:
>
> Thanks for the pointer to Kodkod - it looks very interesting.
>
> I wonder if we aren't talking apples and oranges though. Datalog may
> be a basic reasoner, but it's a simple recursive query langu
Thanks for the pointer to Kodkod - it looks very interesting.
I wonder if we aren't talking apples and oranges though. Datalog may
be a basic reasoner, but it's a simple recursive query language for
data. Can you even get all results out of a SAT solver or do they stop
when satisfiable?
It's for
Yes. I can make a strong endorsement for Kodkod, a Java-based relational
model finder.
http://alloy.mit.edu/kodkod/
I used it fairly extensively last year to solve scheduling problems, and
I've corresponded with its creator.
One problem with Datalog-style reasoners is that, because they want to
g
On Feb 4, 5:22 pm, John Fries wrote:
> Guaranteed-termination is very desirable. However, you can have guaranteed
> termination with an open-world assumption just as well. And I think an
> open-world assumption does a better job of mimicking human reasoning.
>
Do you have a specific reasoner/
Guaranteed-termination is very desirable. However, you can have guaranteed
termination with an open-world assumption just as well. And I think an
open-world assumption does a better job of mimicking human reasoning.
On Wed, Feb 4, 2009 at 2:16 PM, Jeffrey Straszheim <
straszheimjeff...@gmail.com
Well, Datalog does give you guaranteed termination, so there is that,
although its bottom-up strategy is A LOT harder to implement (I'm now
trolling trough about a billion journal articles on "magic sets" and
so on to try to fix this).
I expect to provide full-on evaluable predicates, which I bel
AFAICT, Datalog only supports the closed-world assumption. Does
anyone prefer an open-world assumption reasoner? In my opinion, they
are significantly more powerful.
On Feb 4, 6:16 am, Timothy Pratley wrote:
> > providing relations from clojure-sets and sql-queries.
>
> Wow - this is really ne
> providing relations from clojure-sets and sql-queries.
Wow - this is really neat Erik - thanks for showing
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
"Clojure" group.
To post to this group, send email to cloj
I see. Iris does look pretty good, but I think I'm going to give
writing this in Clojure a try -- the worst outcome is I waste some
time and learn a lot about logic programming. I think Clojure's
superior handling of state and concurrency will pay off here.
On Feb 3, 1:35 pm, hoeck wrote:
> hi
hi jeffrey,
On Feb 3, 2:50 pm, Jeffrey Straszheim
wrote:
> Erik,
>
> Did you use a bottom up evaluation strategy? What top level
> optimizations did you use (e.g. magic sets and so on)?
I only wrote a clojure-wrapper for the iris-reasoner (www.iris-
reasoner.org) mentioned above.
One thing i a
Erik,
Did you use a bottom up evaluation strategy? What top level
optimizations did you use (e.g. magic sets and so on)?
On Feb 3, 6:34 am, hoeck wrote:
> Hi,
>
> On Feb 2, 3:42 pm, Timothy Pratley wrote:
>
>
>
> > Hi Jeffrey,
>
> > On Feb 1, 4:50 am, Jeffrey Straszheim
> > wrote:
>
> > > Ho
Hi,
On Feb 2, 3:42 pm, Timothy Pratley wrote:
> Hi Jeffrey,
>
> On Feb 1, 4:50 am, Jeffrey Straszheim
> wrote:
>
> > However, I'm not sure if you can built your own predicates in Java
> > code (and therefore in Clojure code). That seems like a feature we'd
> > want. I've sent an email to thei
Datalog is a cool problem.
I've started writing some code. The rule-unification part of the
algorithm is trivial -- its not even proper unification at all. The
hard part seems to be optimising the query strategy to avoid
materialising too much. The advantage is you can support rules that
would
It would have been nice if that link was prominent on their website.
They still haven't responded to the email I sent them.
On Feb 2, 10:13 am, Timothy Pratley wrote:
> Just thought I'd share this
> link:http://www.murat-knecht.de/schuerfen/irisdoc/html-single/index.html
> Particularly Example
Just thought I'd share this link:
http://www.murat-knecht.de/schuerfen/irisdoc/html-single/index.html
Particularly Examples 1.2 and 1.6 show how the parts fit together.
I really wish I saw that before attempting anything :) Well now I know
for next time.
--~--~-~--~~~--
Hi Jeffrey,
On Feb 1, 4:50 am, Jeffrey Straszheim
wrote:
> However, I'm not sure if you can built your own predicates in Java
> code (and therefore in Clojure code). That seems like a feature we'd
> want. I've sent an email to their support folks to find out if this
> is possible.
I gave it a
I've been doing some more research on this. This article seems a good
introduction:
http://www.dcc.uchile.cl/~cgutierr/cursos/FDB/p16-bancilhon.pdf
It turns out the naive implementation (a bottom-up fixed point
iterator) is pretty easy to understand, and would not be hard to
implement -- minus
Iris looks really good. They seem to have put a lot of work into
multiple, efficient evaluation strategies, and various levels of
expressivity versus safety. That kind of work is priceless.
However, I'm not sure if you can built your own predicates in Java
code (and therefore in Clojure code).
Doing something like Datalog would be terrific fun. I might
contribute if there is interest.
I'm not an academic, so most of my contributions would be on a
practical level. We'd need someone else to provide the deeper aspects
of theory.
I've read Norvig's book, and understand his code, and my
On Jan 26, 11:44 pm, smanek wrote:
> Hello all,
>
> I'm a Common Lisp programmer who has just started learning Clojure. I
> saw it mentioned online that several members of the existing community
> were looking for someone to build a datalog for Clojure: I was
> wondering what exactly you had in
28 matches
Mail list logo
