Chatlog 2010-11-09

From SPARQL Working Group
Jump to: navigation, search

See original RRSAgent log and preview nicely formatted version.

Please justify/explain all edits to this page, in your "edit summary" text.

14:57:05 <trackbot> Meeting: SPARQL Working Group Teleconference
14:57:05 <trackbot> Date: 09 November 2010
14:57:00 <RRSAgent> Present: LeeF, Souri, pgearon, kasei, cbuilara, AndyS
14:57:22 <LeeF> Agenda:
14:57:24 <LeeF> Chair: LeeF
14:58:00 <LeeF> Regrets: Axel, Alex, Ivan, bglimm, NickH, Olivier, SteveH
15:04:23 <AndyS> scribenick: AndyS
15:04:28 <AndyS> Scribe: Andy Seaborne
15:04:28 <cbuilara> I will add my self to the scribe list
15:04:38 <LeeF> topic: Admin
15:04:39 <LeeF> PROPOSED: Approve minutes at
15:05:57 <LeeF> RESOLVED: Approve minutes at
15:06:18 <LeeF> Next regular meeting: 2010-11-16 @ 15:00 UK / 10:00 EST (Scribe: Chime with NickH as backup)
15:06:34 <LeeF> Regrets next week: LeeF
15:06:59 <AndyS> Partial meeting due to ISWC
15:07:02 <LeeF>
15:09:00 <kasei> q+
15:09:07 <LeeF> ack kasei
15:09:51 <AndyS> Consensus emerging around text in doc : e.g. strict sum using op:numeric-add
15:10:20 <AndyS> q+
15:10:51 <AndyS> q-
15:10:54 <LeeF> for next week: PROPOSED: Close ISSUE-16 by noting that aggregates define their semantics over mixed data types which can include errors, errors that propagate to a SELECT list result in an unbound variable, errors that propagate to GROUP BY clause result in an "unbound" key
15:11:09 <AndyS>
15:12:04 <LeeF> AndyS: formal section will say that syntax -> algebra is 3 steps (instead of current 2)
15:12:08 <LeeF> ... graph patterns, grouping, solution modifiers
15:12:42 <LeeF> ... grouping works over multisets, solution modifiers work over sequences
15:16:04 <pgearon> +q
15:16:22 <LeeF> ACTION: Lee to review the text that contributes to the proposed resolution of ISSUE-16
15:16:22 <trackbot> Created ACTION-332 - Review the text that contributes to the proposed resolution of ISSUE-16 [on Lee Feigenbaum - due 2010-11-16].
15:16:24 <AndyS> ack pgearon
15:16:27 <LeeF> ack pgearon
15:16:49 <AndyS> pgearon: Q about differing types for sum: 
15:17:53 <AndyS> LeeF: defn using op:numeric-add => error to add non-number.  
15:18:22 <AndyS> ... when used in SELECT expressions erro=> unbound (extend operator)
15:19:38 <pgearon> q-
15:19:48 <LeeF> pgearon: not clear from text what happens when an error is encountered midway through a SUM evaluation
15:20:17 <LeeF> topic: update shortcuts
15:20:18 <LeeF>
15:21:12 <AndyS> LeeF: concerns at shortcuts this early in life of SPARQL 1.1
15:22:14 <AndyS> ... Shortcuts: ADD, MOVE, COPY
15:22:39 <kasei> I'm mildly opposed to the shortcuts. Same concerns as LeeF.
15:22:44 <AndyS> ... analogs of the file operations and similar to REST operations.
15:22:46 <Souri> ADD => is the semantics like a MERGE? b/c we may have duplicates
15:22:56 <Souri> q+
15:23:01 <LeeF> ack Souri
15:24:15 <AndyS> AndyS: Usage experience - quickly started using them after adding them for remote admin of datasets.
15:24:18 <AndyS> Strawpoll:
15:24:26 <LeeF> straw poll on proposed ADD/COPY/MOVE update shortcuts (+1/0/-1):
15:24:31 <AndyS> +1
15:24:32 <LeeF> -1
15:24:33 <kasei> -1
15:24:33 <Souri> +1
15:24:34 <pgearon> 0
15:25:25 <cbuilara> +1
15:26:24 <Souri> q+
15:26:40 <LeeF> LeeF: I am motivated a bit by hearing positive implementation and usage experience from AndyS
15:26:43 <LeeF> ack Souri
15:27:23 <LeeF> Souri: easier to spot optimizations for the shortcut operations
15:28:00 <LeeF> summary: slight preference in today's group for including shortcuts - motivation is ease of use of language, ease of optimization
15:28:17 <LeeF> topic: grouping by expressions
15:28:27 <AndyS>
15:28:45 <LeeF> (btw, the last conversation dealt with ISSUE-59)
15:29:22 <AndyS> (no issue for the issue of GROUP BY expression)
15:29:45 <AndyS> LeeF summary/proposal possibilities: 0187
15:30:23 <AndyS> ... WG wanted to do in principle - mechanism postponed last time.
15:31:00 <LeeF> option 1 - To project grouping expressions, use BIND to alias the expression to 
15:31:00 <LeeF> a variable and then GROUP BY and project that variable (as above).
15:31:25 <LeeF> option 2 - Include an AS aliasing mechanism in GROUP BY, allow that alias to be 
15:31:25 <LeeF> projected in the SELECT clause
15:32:34 <AndyS> Discussion around opt 2 and opt 1.
15:32:49 <LeeF> q?
15:32:59 <AndyS> SteveH noted wish for if opt 2, then allow no AS form as well.
15:33:00 <Souri> q+
15:33:07 <LeeF> ack Souri
15:33:48 <AndyS> Souri: Q about BIND and GROUP BY
15:33:55 <LeeF> SELECT ... ?x ... WHERE { .... BIND ( ?x AS expr ) } GROUP BY ?x
15:34:18 <kasei> that BIND is backwards, no?
15:34:25 <LeeF> SELECT ... ?x ... WHERE { .... } GROUP BY (expr AS ?x)
15:34:39 <LeeF> should have been BIND ( expr AS ?x )
15:35:51 <AndyS> SELECT .... WHERE { .... } GROUP BY (expr AS ?x)
15:36:00 <AndyS> SELECT .... WHERE { .... } GROUP BY (expr)
15:36:44 <LeeF> (general consensus against _requiring_ alias when grouping by expressions)
15:36:59 <LeeF> straw poll: allow option 2 (include AS aliasing in GROUP BY syntax) (+1/0/-1)
15:37:49 <LeeF> straw poll: allow option 2 (include expressions and AS aliasing in GROUP BY syntax) (+1/0/-1)
15:37:58 <kasei> +1
15:38:00 <LeeF> 0
15:38:04 <AndyS> +1
15:38:13 <Souri> +1
15:38:20 <cbuilara> 0
15:38:25 <pgearon> 0
15:39:43 <LeeF> for next week: PROPOSED: The GROUP BY clause can include expressions with or without AS aliases. These aliases can be re-used in the HAVING clause and query result clause
15:39:46 <pgearon> I don't see the need for it when a BIND in the WHERE clause will work just as well. But I don't have a particular problem with it, so I voted 0 instead of -1
15:39:48 <AndyS> leeF: is teh ?x usable in HAVING?
15:39:51 <AndyS> AndyS; yes
15:40:00 <Souri> q+
15:40:10 <AndyS> Topic: Function library
15:40:29 <LeeF> ack souri
15:42:10 <AndyS> group - extend - having
15:43:00 <AndyS> LeeF: will discuss in full group
15:43:06 <pgearon> +q
15:43:11 <LeeF> ack pgearon
15:45:39 <kasei> also, specifying an algorithm would prevent implementations from doing something better (like sourcing from :)
15:45:41 <pgearon> -q
15:46:25 <pgearon> I'd prefer not to see a specified algorithm, but I wanted to raise it because it results in an untestable feature
15:46:41 <AndyS> Topic: test cases
15:46:44 <kasei> I believe I'm passing all the non-entailment tests right now
15:47:00 <AndyS> LeeF: need to keep building test case materials
15:47:09 <AndyS> ... but currently on track
15:47:41 <AndyS> ADJOURNED
15:53:29 <Zakim> SW_(SPARQL)10:00AM has ended
15:53:31 <Zakim> Attendees were LeeF, Souri, pgearon, kasei, cbuilara, AndyS