RDB2RDF Working Group Teleconference -- 28 Feb 2012

<trackbot> Date: 28 February 2012

<Ashok> meeting: RDB2RDF

macted, nunolopes, aaaa isn't actually us

<MacTed> d'oh!

<scribe> scribenick: cygri

<scribe> scribe: Richard Cyganiak

Admin

<Ashok> PROPOSAL: Accept the minutes of last meeting http://www.w3.org/2012/02/14-RDB2RDF-minutes.html

<MacTed> cygri - aaaa was the line Zakim recognized when your call pinged in... *shrugs*

<boris> +1

RESOLUTION: Accept the minutes of last meeting http://www.w3.org/2012/02/14-RDB2RDF-minutes.html

Party

ashok: we are now officially CR. thank you all for the hard work. this is a significant step
... specs are considered complete and correct. focus on test cases and implementation report

Implementation report and Test Cases

ashok: test cases should be complete by the end of this month
... can we get an update?

boris: i improved the RDFa markup in TC documents
... got some feedback from Juan to update some test cases
... i added some test cases on R2RML data errors
... i need to add more test for data errors
... group needs to check the current set

juansequeda: i went through all test cases
... our tool now passes all of them
... i want to see more data errors and tests where mappings are wrong
... our charter says something about making R2RML mappings for schemas like wordpress
... we only have very small examples now
... we should do something on larger schemas, maybe northwind
... i know this would imply more work

<ericP> i wonder if there's any way to get some interested users to create a note for e.g. northwind

<ericP> cygri: need to define how good the test cases have to be to give us confidence to go to REC

<juansequeda> +q to ask how are we on coverage

<ericP> ... don't need to cover everything, but we want to demonstrate interop

<ericP> ... most important input is from implementors

<ericP> ... so far, i've been eyeballing the tests

<ericP> ... while juan has been running the tests

cygri: i'm inclined to listen to feedback from implementers. if they think we need more tests, then we should probably make more

ivan: things like the wordpress mapping shouldn't be put on the critical path
... but would be great if someone could do it and demonstrate it

<juansequeda> I can work on the wordpress r2rml

<Ashok> Juan, that would be great!

boris: we are time constrained. need feedback from implementers mostly

<boris> http://www.w3.org/2001/sw/rdb2rdf/wiki/R2RML_TC

<MacTed> http://www.w3.org/2001/sw/rdb2rdf/wiki/R2RML_TC

juansequeda: how are we doing on R2RML test coverage?

boris: we cover most if not all language features. data errors not yet so well

juansequeda: i'd like to see test cases for weird corner cases that may be unexpected but are allowed

<Zakim> juansequeda, you wanted to ask how are we on coverage

ericP: someone could do northwind etc when we're languishing in PR and do as WG Note
... i agree that weird corner cases make good test cases. in sparql, we always encoded questions "does it do this or that" as a test case

<ericP> cygri: +1 to corner cases

<Ashok> Richard: Would be good to have some weird corner case tests

<ericP> ... how do we generate them?

<ericP> ... when folks ask questions, we can capture those in test cases

<ericP> ... i can think of a few corners of the spec

<ericP> ... is there anything we can do to cover these edges better?

juansequeda: predicate-object maps can have multiple predicate and object maps. that's a bit confusing. i will write an email on that

<Zakim> MacTed, you wanted to ask whether matrix was built from TCs, or TCs built to fill matrix

MacTed: are the test cases built to fill the matrix, or is the matrix extended to document the test cases?

boris: the matrix came first

MacTed: so where do the negative test cases fit on the matrix?

ericP: conformance testing is not our job. we need to demonstrate that well-meaning implementers can understand the spec and produce interoperable specs
... nevertheless, in SPARQL we went pretty far down the path of systematically covering everything
... just capturing questions didn't suffice to fill in the matrix

<ericP> cygri: as with HTML5, every conformance statement can beget a test

<Souri> minor spelling error: In the Matrix, replace "Foreing" with "Foreign"

<ericP> +1 to thanking boris

juansequeda: thanks boris for this hard work!

+1 to that!

<Ashok> +1

<ericP> .RESOLVED: thanks boris

<MacTed> http://troels.arvin.dk/db/rdbms/#functions-concat

juansequeda: how do we deal with vendor dialects in the tests?

<boris> so, || is the standard

<Ashok> Ted dsicussses the link he posted above

<Souri> From a website: The concatenation operator used by Oracle is ||, which is also ANSI-compliant.

<ericP> cygri: where possible, we should write R2RML mappings which work in as many places as possible

<nunolopes> I previously ran the tests with mysql by changing server properties

<nunolopes> as follows:

<nunolopes> [mysqld]

<nunolopes> sql-mode="PIPES_AS_CONCAT,ANSI_QUOTES"

<nunolopes> default-storage_engine=innodb

<nunolopes> might not be desirable

<ericP> ... for the cases where that's not possible, e.g. concatonation, we could write instructions with search and replace to produce vendor-specific SQL

<ericP> ericP: not easy for concat

<dmcneil> +q

<boris> q to ask Juan if || is the only vendor specific issue?

<ericP> cygri: we only have to cover the [syntax of the] tests which we've written

juansequeda: will we freeze the test cases and make a zip?

<Ashok> Boris: There is a zip file with the test cases

juansequeda: some tests use fixed schema name. that's difficult for running the tests, we can't assume that the user has this schema

boris: i don't remember why we have that. i could remove it

<nunolopes> +1

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0002d

<juansequeda> Select ('Student' || "ID" ) AS StudentId

<juansequeda> , "ID"

<juansequeda> , "Name"

<juansequeda> from "SA.Student"

<juansequeda> CREATE TABLE "Student" (

<juansequeda> "ID" INTEGER,

<juansequeda> "Name" VARCHAR(15)

<juansequeda> );

<juansequeda> INSERT INTO "Student" ("ID", "Name") VALUES(10,'Venus');

<Souri> we need to have test cases with explicit schema name and table name

<MacTed> fix *appears* to be replacing all TestCase instances of "SA.Student" with "Student"

<MacTed> table creation scripts need to be checked

<Zakim> cygri, you wanted to say just do hg clone

<ericP> cygri: re: freezing tests or downloading them as a zip, i think it would be reasonable to point to the hg repo

<dmcneil> I think the test cases should have queries that comply with SQL 2008, and these queries should be identified as being http://www.w3.org/ns/r2rml#SQL2008

<ericP> ... if folks want to use them locally, they can $(hg clone http://dvcs.w3.org/hg/rdb2rdf-tests/)

<ericP> +1 to using hg

<boris> +1 to keep both

Souri: there should be more variation in test cases, e.g., quote some column identifiers and don't quote others
... also, there should be at least one test case that has a schema name to make this clear

ericP: while we're still changing tests, we should tell people to use mercurial to reduce the risk of them running outdated tests
... we can freeze a zip at the end

http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0002d

<ericP> cygri: looking at <above>, some of the IDs are quoted and some are not

<ericP> ... id and name are quoted while studentid is not

rr: column "ID"
... column "\"ID\""

<Zakim> ericP, you wanted to ask if i may update <https://dvcs.w3.org/hg/rdb2rdf-tests/raw-file/tip/D017-I18NnoSpecialChars/directGraph.nt>

<ericP> s/_:a <成分#植物名.使用部> <植物/名-しそ.使用部-葉> .

<ericP> /_:a <成分#ref-植物名.使用部> <植物/名-しそ.使用部-葉> ./

<Souri> that looks chinese to me

<ericP> in fact, it's CLK

cygri: if identifiers are quoted in the CREATE TABLE statement or SQL query, then they also need to be quoted in the R2RML literal, so we end up with "\"xxx\""

<juansequeda> Souri, are you talking about R2RMLTC0002b

<ericP> cygri: this is yet another corner case (note for tests)

<ericP> ... the SQL standard says that:

<ericP> ... .. quoted identifiers are case-sensitive

<ericP> ... .. unquoted identifiers are case-insenstive

cygri: ID == Id, "ID" != "Id", "ID" == ID, "ID" == Id

ericP, it's standard SQL

<ivan> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2012Feb/0085.html

ericP, see here for some details: http://www.w3.org/TR/r2rml/#dfn-column-name