13:56:24 <RRSAgent> RRSAgent has joined #sparql
13:56:24 <RRSAgent> logging to http://www.w3.org/2010/05/27-sparql-irc
13:56:27 <LeeF> RRSAgent, make logs public
13:58:35 <Zakim> Team_(sparql)13:55Z has now started
13:58:41 <Zakim> +Lee_Feigenbaum
13:59:18 <AndyS> Just skyping up ...
13:59:32 <Zakim> +kasei
14:00:21 <Zakim> +[IPcaller]
14:00:35 <AndyS> zakim, [IPcaller] is me
14:00:35 <Zakim> +AndyS; got it
14:01:23 <pgearon> I'm being told that the code 77277# is not valid
14:01:30 <MattPerry> MattPerry has joined #sparql
14:01:30 <LeeF> it's 772775# today
14:01:38 <pgearon> thanks
14:01:41 <LeeF> yup
14:01:46 <pgearon> I see it now
14:01:48 <LeeF> zakim, code?
14:01:51 <Zakim> the conference code is 772775 (tel:+1.617.761.6200 tel:+33.4.89.06.34.99 tel:+44.117.370.6152), LeeF
14:01:58 <Zakim> +pgearon
14:03:09 <Zakim> +MattPerry
14:04:07 <AndyS> http://www.w3.org/2009/sparql/docs/property-paths/Overview.xml#Outstanding_Issues
14:04:17 <AndyS> esp. duplicates.
14:04:32 <Zakim> +OlivierCorby
14:04:39 <LeeF> Scribenick: LeeF
14:04:43 <pgearon> I'll take it over
14:05:15 <kasei> and I've got an issue/question to add to the list
14:05:19 <LeeF> AndyS: most important issue is the cardinality of results
14:06:22 <LeeF> ... other issues listed at http://www.w3.org/2009/sparql/docs/property-paths/Overview.xml#Outstanding_Issues
14:08:23 <LeeF> kasei: 0-length paths -- what are the possible nodes that can match the ends of that path?
14:10:18 <LeeF> ... if you have 2 variables on the end of a 0-length path, what do they match? document suggests infinite solutions
14:10:31 <LeeF> ACTION: Greg to send test cases for 0-length paths to mailing list
14:10:31 <trackbot> Created ACTION-251 - Send test cases for 0-length paths to mailing list [on Gregory Williams - due 2010-06-03].
14:10:45 <kasei> Zakim, mute me
14:10:45 <Zakim> kasei should now be muted
14:11:10 <LeeF> http://lists.w3.org/Archives/Public/public-rdf-dawg/2010AprJun/att-0275/Property_Path_Diagram.pdf
14:11:51 <LeeF> AndyS: in diagram 1, there are 2 simple paths from A to Z
14:11:58 <LeeF> ... no cycles in it
14:12:31 <AndyS> :A :p{2} ?Z :: how many answers
14:13:03 <LeeF> Do you get ?Z=:Z once or twice?
14:13:17 <AndyS> { :A :p ?X . ?X :p ?Z }
14:13:32 <LeeF> ... if you wrote that graph pattern explicitly, you'd get 2 answers
14:13:59 <pgearon> q+
14:14:13 <LeeF> ack pgearon
14:14:35 <LeeF> pgearon: the way SPARQL works now, i expected to get 2 results out of this
14:14:48 <LeeF> ... i used to do set thinking everywhere, but had to change that thinking to be SPARQL-compliant
14:15:03 <LeeF> ... according to SPARQL thinking, I would really expect to see 2 results come out of this even though the extra variable is projected away
14:15:10 <AndyS> zakim, who is on the phone?
14:15:10 <Zakim> On the phone I see Lee_Feigenbaum, kasei (muted), AndyS, pgearon, MattPerry, OlivierCorby
14:15:18 <LeeF> { :A :p [ :p ?Z ] } also has 2 results
14:16:11 <LeeF> AndyS: think it's not that hard to implement because a breadth-first search will get you the right answers 
14:17:52 <LeeF> <discussion of wild life>
14:18:12 <AndyS> Diagram 2: { :A :p{2} ?Z }
14:18:42 <LeeF> AndyS: in diagram 2, the same pattern ...
14:18:48 <LeeF> ... you get the same answers form diagram 1
14:18:58 <LeeF> ... but there's also A->C->C
14:19:16 <LeeF> ... so there is a solution with ?Z=:C ?
14:19:53 <LeeF> AndyS: ?Z=:C is not a simple path solution, because you've gone around a loop
14:20:03 <LeeF> pgearon: but no potential to get confused between paths?
14:20:32 <LeeF> AndyS: using technical sense of "simple path" from graph theory
14:21:19 <bglimm> bglimm has joined #sparql
14:21:30 <AndyS> http://en.wikipedia.org/wiki/Graph_%28mathematics%29
14:21:36 <pgearon> http://en.wikipedia.org/wiki/Path_(graph_theory)
14:22:34 <LeeF> AndyS: I think there should be 3 solutions, ?Z = :Z, :Z, :C
14:22:49 <pgearon> +1 to 3 solutions
14:22:56 <AndyS> Diagram 2: { :A :p{3} ?Z }
14:23:23 <LeeF> ... excludes the A->B->Z route
14:23:34 <LeeF> A->C->C->C
14:23:38 <LeeF> A->C->C->Z
14:24:23 <LeeF> pgearon: I get into trouble here because when I see a loop I start dropping off
14:25:04 <AndyS> Diagram 2: { :A :p* ?Z }
14:25:11 <AndyS> Diagram 2: { :A :p* ?end }
14:26:14 <AndyS> { :A :p{3} ?Z } == { :A :p ?e . ?e :p ?f . ?f :p ?end }
14:26:25 <LeeF> What happens if we say that fixed-length property paths have exact cardinality requirements but variable-length paths (* or +) have distinct cardinality semantics?
14:27:50 <MattPerry> I like Lee's idea
14:29:06 <LeeF> AndyS: sounds like a reasonable scheme, we should do it by defining the cardinality for each operator
14:29:06 <AndyS> elt{n,} 
14:29:25 <AndyS> elt{n,m}
14:30:30 <pgearon> LeeF: expect that patterns that can be viewed as a shortcut for a fixed length expression should have the same cardinality as the fully expended expression
14:31:05 <AndyS> Diagram 2: { :A :p* ?Z } 
14:31:06 <LeeF> PROPOSED: Fixed-length path solutions should be the same as their expansion in triples with all variables projected; variable-length paths have distinct cardinality semantics
14:31:12 <pgearon> LeeF: patterns that have an unbounded length with have a cardinality based on ignoring repeated nodes that appear in loops
14:32:08 <LeeF> Not distinct as in "only one ?Z"... distinct as in "don't add more from following loops"
14:32:14 <pgearon> {:A :p{2,} ?Z}
14:32:15 <LeeF> AndyS: not sure how to say that
14:32:44 <AndyS> Diagram 2: {:A :p{4,} ?end }
14:33:25 <MattPerry> How about number of paths that do not repeat a node?
14:33:59 <AndyS> { :A :p{3} ?Z } == { :A :p ?e . ?e :p ?f . ?f :p ?end }
14:34:18 <LeeF> MattPerry: use the criteria of not repeating a node for the unbounded cases
14:34:29 <LeeF> AndyS: so finite ones would still be expansion into alternatives
14:34:49 <AndyS> :p{2,3} =?= :p :p :p?
14:35:30 <LeeF> elt{n,}
14:35:43 <LeeF> elt{n}
14:35:50 <AndyS> :p{2,} =?= :p :p :p*
14:36:20 <LeeF> card(elt{n,}) is card(elt{n}) + card(paths longer than n that don't repeat a node) ?
14:36:49 <kasei> i don't envy having to come up with spec text for that...
14:36:50 <pgearon> { :A :p ?Z }  =  { { :A :p ?e . ?e :p ?Z } UNION { :A :p ?e . ?e :p ?f . ?f :p ?Z } }
14:37:28 <LeeF> s/:p/:p{2,3}
14:37:46 <pgearon> corrected to:  { :A :p{2,3} ?Z }  =  { { :A :p ?e . ?e :p ?Z } UNION { :A :p ?e . ?e :p ?f . ?f :p ?Z } }
14:38:50 <kasei> lhs results = Z, Z, C
14:38:55 <AndyS> LHS: A-B-Z, A-C-Z, A-C-C
14:39:00 <LeeF> card(elt{n,m}) = SUM(i = n to m) of card(elt{i}) ?
14:39:26 <kasei> sounds right
14:39:35 <AndyS> RHS: A-C-C-Z, A-C-C-C
14:40:24 <LeeF> This may be a new motivation for people to start using REDUCED
14:40:27 <LeeF> :)
14:42:39 <pgearon> { :A :p{2,} ?Z }  =  { { :A :p ?e . ?e :p ?Z } UNION { :A :p+ ?Z MINUS { { :A :p ?Z } UNION { :A :p ?Z } } }
14:43:12 <pgearon> { :A :p{2,} ?Z }  =  { { :A :p ?e . ?e :p ?Z } UNION { :A :p+ ?Z MINUS { :A :p ?Z } } }
14:43:28 <kasei> Zakim, unmute me
14:43:28 <Zakim> kasei should no longer be muted
14:43:40 <LeeF> AndyS: MINUS messes up cardinality
14:45:47 <pgearon> I had originally thought of :p{2,} as being :p+ minus the 1-step case
14:46:10 <LeeF> PROPOSED: The cardinality of solutions to fixed-length paths is the same as the cardinality of solutions to the path expanded into triple patterns (with all variables projected); the cardinality of solutions to variable-length paths is the cardinality of solutions via paths that do not repeat nodes; the cardinality of solutions to paths combining fixed anv ariable length (elt{n,} ) is a combination of the fixed definition plus the variable definition for paths long
14:46:27 <pgearon> but the new approach has me thinking that the :p{2,} as being a mix between two types of cardinality rules
14:47:50 <pgearon> { :A :p{2,} ?Z }  =  { { :A :p ?e . ?e :p ?Z } UNION { :A :p+ ?Z MINUS { { :A :p ?Z } UNION { :A :p ?e . ?e :p ?Z } } }
14:50:30 <LeeF> zakim, who's on the phone?
14:50:30 <Zakim> On the phone I see Lee_Feigenbaum, kasei, AndyS, pgearon, MattPerry, OlivierCorby
14:51:25 <kasei> { :A :p{2,} ?Z }
14:51:37 <kasei> { :A :p+ ?Z MINUS { { :A :p ?Z } UNION { :A :p ?e . ?e :p ?Z  } } }
14:51:55 <kasei> { :A :p ?Z }
14:52:11 <AndyS> { :A :p{2,} ?Z }  ==> { :A :p/:p ?x . ?x :p* ?Z } ??
14:53:35 <kasei> :p/:p is just :p{2}?
14:53:58 <kasei> A->C->Z->Q
14:53:58 <kasei> A->Q
14:55:04 <Zakim> This conference is in overtime; all ports must be freed
14:55:45 <LeeF> zakim, extend conference 
14:55:45 <Zakim> I don't understand 'extend conference', LeeF
14:56:00 <LeeF> zakim, 60 more minutes
14:56:00 <Zakim> I don't understand '60 more minutes', LeeF
14:56:25 <kasei> :p+ will result in bindings C, Z, Q, Q
14:58:21 <Zakim> -Lee_Feigenbaum
14:59:05 <OlivierCorby> what about (p1/p2/p3)* where p1/p2 stay on same node ?
15:00:42 <AndyS> (:p{3})*
15:02:13 <kasei> issue is whether repeated nodes restriction affects internal nodes of (p1/p2/p3) when considering a pattern of (p1/p2/p3)*
15:02:19 <AndyS> (p1/p2/p3)
15:02:50 <MattPerry> I would say treat (p1/p2/p3) as a unit and don't repeat that
15:03:28 <kasei> AndyS: best approach is to provide more detailed description and try test cases
15:04:11 <kasei> next issue
15:04:17 <kasei> Zakim, mute me
15:04:45 <AndyS> ?a !(rdf:type) ?b
15:04:46 <Zakim> kasei should now be muted
15:05:11 <AndyS>       ?x ! ( rdf:type | ^rdf:type)* ?y
15:05:55 <pgearon> AndyS: can I find a connection between two individuals, but I don't want to find the connection by going up the class hierarchy
15:06:07 <kasei> what is the issue here?
15:06:47 <pgearon> AndyS: issue is complexity of implementation
15:06:52 <AndyS> ?x (!( rdf:type | ^rdf:type)) * ?y
15:07:29 <AndyS> [^A-Z]
15:08:58 <kasei> yeah, seems reasonable
15:09:14 <AndyS> !( rdf:type | ^rdf:type) 
15:09:21 <AndyS> !(...)
15:09:26 <pgearon> AndyS: Seems like a reasonable thing to want. Will people consider it for possible problems?
15:09:50 <kasei> chaining of the unbounded modifiers scares me: (!(:p+))*
15:10:07 <kasei> Zakim, unmute me
15:10:07 <Zakim> kasei should no longer be muted
15:10:43 <kasei> and we consider ^:p as a single thing?
15:10:57 <pgearon> Can it be re-expressed as:  { ?x (! rdf:type)* ?y } UNION { ?y (! rdf:type)* ?x }   ???
15:11:37 <kasei> probably not if you combine it with more complex paths
15:13:22 <kasei> :X :type :Y ^:type :Z
15:13:36 <MattPerry> +1 for this
15:14:14 <kasei> I think the analog to [^abc] is the right guide here. just atomic things inside the negation.
15:14:31 <MattPerry> +1 to kasei
15:14:41 <pgearon> +1 to kasei
15:14:49 <kasei> well, atomic with |
15:15:21 <pgearon> AndyS: will write a proposal for that
15:16:21 <AndyS> PROPOSED: Add negated property class (c.f regex [^abc]) where the negated property set is a a distjunction of fwd and backward properties only
15:17:31 <AndyS> \
15:17:47 <kasei> ☃
15:17:54 <kasei> return of the snowman
15:17:57 <AndyS> ~
15:19:14 <pgearon> AndyS: next issue - the one about lengths
15:19:19 <kasei> is there a syntax proposal for that?
15:20:10 <MattPerry> I don't know how useful the length is without the actual path
15:20:25 <pgearon> AndyS: gets more complex with syntax, and also now you have to consider the actual path taken
15:21:47 <kasei> pgearon: can't see how useful it is without an actual path description, which you could get the length from directly.
15:22:03 <AndyS> PROPOSED: time constraints on the WG mean we will not support path length in this cycle.
15:22:17 <MattPerry> +1 for that
15:22:20 <pgearon> +1
15:22:20 <AndyS> PROPOSED: time constraints on the WG mean we will not support path length in this working group
15:23:19 <pgearon> AndyS: only way to consider this would be if someone showed up with a concrete proposal on how this is to be done
15:23:56 <kasei> well, or it works if we have path lengths :)
15:24:27 <pgearon> AndyS: new issue - results ordered by list
15:25:13 <pgearon> AndyS: ordering is a difficult issue since it won't be preserved through the rest of the query execution
15:25:39 <pgearon> AndyS: knows that some people have been relying on the fact that lists are returned in order in his implementation, but they've been lucky so far
15:26:12 <pgearon> pgearon: to preserve ordering we need some kind of "list" object in results
15:26:27 <AndyS> PROPOSED: WG will not consider ordering nor preserving ordering or results from property paths
15:26:46 <kasei> s/or/of/ ?
15:27:01 <bglimm> bglimm has joined #sparql
15:27:10 <kasei> +1
15:27:14 <AndyS> +1
15:27:18 <MattPerry> +1
15:27:21 <pgearon> AndyS: for more complicated expressions it gets too hard to ensure that ordering is preserved
15:27:23 <pgearon> +!
15:27:26 <pgearon> +1
15:28:46 <bglimm> bglimm has left #sparql
15:29:07 <bglimm> bglimm has joined #sparql
15:29:13 <bglimm> bglimm has left #sparql
15:29:26 <bglimm> bglimm has joined #sparql
15:29:39 <AndyS> ?x :p^:q ?y
15:29:47 <AndyS> ?x :p/^:q ?y
15:29:53 <kasei> no. prefer unary /^
15:30:04 <MattPerry> +1 for unary
15:30:07 <pgearon> AndyS: is anyone keen to have a binary hat as opposed to unary (examples above)
15:30:26 <OlivierCorby> +1 for unary
15:30:32 <pgearon> +1 for unary
15:30:49 <AndyS> PROPOSED:  Only unary ^, no binary ^
15:30:58 <kasei> +1
15:31:01 <MattPerry> +1
15:31:05 <pgearon> +1
15:31:21 <OlivierCorby> +1
15:31:28 <AndyS> ?x ^:q ?y
15:31:47 <AndyS> x :p/^:q ?y
15:32:16 <AndyS> ?x :p^:q ?y
15:33:43 <AndyS> Doc says :: elt1 ^ elt2	 Shorthand for elt1 / ^elt2, that is elt1  followed by the inverse of elt2.
15:34:32 <OlivierCorby> does ^ apply to property only or also to expression
15:35:35 <pgearon> pgearon: since ^ is just a convenient shorthand for /^ then I would like the option of dropping the /
15:36:40 <AndyS> ^uri
15:36:53 <AndyS> Maybe only allow ^uri, not ^expression
15:37:20 <kasei> I'd prefer the ^uri form only, but haven't thought out the implications.
15:37:23 <MattPerry> +1 for only ^uri
15:37:30 <AndyS> ?x ^expr ?y == ?y expr ?x
15:37:40 <pgearon> I'd like ^?var
15:38:09 <kasei> except the ^ doesn't have to be the outtermost operator
15:40:16 <pgearon> AndyS: thinks that it's OK, since once you get to the ^ part then you just have to change the direction around at that point
15:40:41 <MattPerry> ^(p1/p2/p3) == (p3/p2/p1) ?
15:41:17 <kasei> right, so that's not just switching the endpoint nodes. that's flipping the insides of the expression.
15:42:08 <AndyS> ?x ^(p1/p2/p3) ?y   ::  ?y p1/p2/p3 ?x :: ?x ^p3/^p2/^p1 ?y
15:44:04 <pgearon> AndyS: all 3 of these should be equivalent
15:47:31 <pgearon> AndyS: new issue: relationship with inference
15:47:57 <pgearon> AndyS: can get nasty with more than one part of the system trying to do transitivity
15:48:00 <kasei> I don't understand the issue.
15:48:41 <bglimm> Zakim, passcode?
15:48:41 <Zakim> the conference code is 772775 (tel:+1.617.761.6200 tel:+33.4.89.06.34.99 tel:+44.117.370.6152), bglimm
15:49:08 <kasei> is the issue cardinality?
15:49:43 <AndyS> Oops:: bglimm, we don't know how to fix it.
15:49:52 <kasei> Zakim thinks our call ended 50 minutes ago...
15:50:03 <AndyS> and we are about to be evicted anyway I guess
15:50:38 <AndyS> We will take the entailment interaction to the list for ENT+PP discussion
15:50:44 <bglimm> ok
15:51:09 <bglimm> I can try and ask Zakim for a room?
15:51:24 <bglimm> Zakim, room for 10?
15:51:26 <Zakim> ok, bglimm; conference Team_(sparql)15:51Z scheduled with code 772775 (SPARQL) for 60 minutes until 1651Z
15:52:32 <bglimm> Hm, doesn't work, I'll follow the list then
15:53:47 <kasei> is the doc even defined with vars right now?
15:54:54 <AndyS> not currently
15:55:54 <AndyS> ADJOURNED
15:56:05 <Zakim> -kasei
15:56:09 <Zakim> -MattPerry
15:56:10 <pgearon> talk to you all Tuesday
15:56:11 <Zakim> -AndyS
15:56:13 <Zakim> -OlivierCorby
15:56:17 <Zakim> -pgearon
15:56:19 <Zakim> Team_(sparql)13:55Z has ended
15:56:21 <Zakim> Attendees were Lee_Feigenbaum, kasei, AndyS, pgearon, MattPerry, OlivierCorby
16:16:39 <AndyS> rrsagent, make minutes public
16:16:39 <RRSAgent> I'm logging. I don't understand 'make minutes public', AndyS.  Try /msg RRSAgent help
16:17:33 <AndyS> OK - which bot makes the IRC log public?
16:45:39 <AndyS> Ah - already public.  Done.
16:53:08 <AxelPolleres> AxelPolleres has joined #sparql
17:12:05 <AxelPolleres> AxelPolleres has left #sparql
17:48:45 <AxelPolleres> AxelPolleres has joined #sparql
19:53:30 <bglimm> bglimm has joined #sparql
20:47:47 <AndyS> AndyS has joined #sparql
23:02:39 <LeeF> LeeF has joined #sparql