RDB2RDF Working Group Teleconference

<trackbot> Date: 05 October 2010

<Souri> Hi Pat

<Angela> maybe is me...i am using gmail...

<mhausenblas> scribenick: nunolopes

<mhausenblas> juansequeda shall we wait for you?

<juansequeda> ok

admin

<mhausenblas> PROPOSAL: Accept the minutes of last meeting, see http://www.w3.org/2010/09/28-rdb2rdf-minutes.html

RESOLUTION: minutes of last meeting approved

<mhausenblas> Review open actions, see http://www.w3.org/2001/sw/rdb2rdf/track/actions/open

<mhausenblas> ACTION-73?

<trackbot> ACTION-73 -- Michael Hausenblas to clean up the actions -- due 2010-10-05 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/73

FPWD of mapping language

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Oct/att-0002/rdb2rdf_fpwd.html

<Seema> yup...it came out inline..i have sent you the document..can you email it as an attachment

<hhalpin> just wondering if we want to put that in W3C CVS space and do a pubrules check.

Souri: there is a new version of the draft, sent by seema

<hhalpin> that would stop the e-mail problems :)

Souri: we can work on this version for now
... we still didn't manage to get W3C CVS working

… we used a template for the document

hhalpin: I can help you to put it online after the telco

<juansequeda> hhalpin: anyway you can contact the w3c operator so they can connect me to the call? That is how I usually do it but I guess they are on a coffee break?

mhausenblas: we use the available version for now

<hhalpin> Juan: what's the issue? voip (skype?) not working?

<mhausenblas> Michael: use : http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Oct/att-0002/rdb2rdf_fpwd.html for the time being

<juansequeda> voip

<juansequeda> never works

Souri: we have overview and intro sections

… R2ML specified using Turtle syntax

… instead of calling classMap we called table2triplesMap

… de point is to simplify for the reader

… if table can be mapped to logical triples this would be a better name

<juansequeda> yes

… instead of calling things as instanceId what we are getting is subjects and property object pairs (?)

<cygri> mhausenblas, I confused zakim by telling him that "/me" is with you ... does this need to be fixed?

<mhausenblas> no worries ;)

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Oct/att-0004/rdb2rdf_fpwd.html

… this latest version has an updated example

scribe: Section 3.1

mhausenblas: before that, I asked to define the intended audience for the doc, is that done?

Souri: no, not yet

… we created a class hiearchy

… rdfTermMapClass

…. with subclasses for IRI, bnode, literal

… we defined all the properties that are relevent in the rdfTermMap

… 3.1.1 rr:property

… identifies the property name that will be used

…. 3.1.1.2 rr:column gives us the database column name

… 3.1.1.3 rr:inverseExpression

mhausenblas: maybe there is no need to go through all the classes and properties

Souri: the properties are important to understand section 3.3

mhausenblas: maybe it's more productive to decide how to do with the document

hhalpin: we can keep going though the document and then work on mailing list

<hhalpin> just noting we can ask for FPWD next week.

Souri: 3.1.1.4 specifies the datatype of a literal

… similar for language tag in 3.1.1.5

… rr:columnGraph you can assing a particular triple to a specific graph

… could also allow multiple graphs

…. rr:constantValue 3.1.1.7

… for all the rows, the value of that property shoudl be a constant

… rr:isRDFType says there could be many triples as IRI

… some of then are treated as rdf:type

… rr:isComputed allows the property name to be defined by another column

… instead of being a constan

… 3.2 is simple

… Main section is 3.3 Table to triplesMap

… should be renamed to triplesMap

… tells how to formulate the triples for a particular row

… it's a rdfs:class represented by an SQL query

??: what are the elements of this class?

<hhalpin> an instance of the class is a particular mapping, correct?

Souri: this is actually a description of a class,

hhalpin: what is it, a piece of text, … ?

Souri: example, you have a classs and you want to create a mapping object that belongs to the class

… this needs to have a set of properties inside

… how to get subject, prop, object

… on sec 3.3.1 there is an example

… the idea is that the obejct is a description of the class

… describes how we go about describing the classes

… Section 3.3.1

… rr:logicalTable can be a query or a name of a table

… cardinality must be 1

… rr:class : the instances will be generated as members of this class

… can be optional, but there are some implications

… rr:tableGraphMap property : at the logical table level, all the triples that would be generated should belong to the specified named graph

… rr:subjectMap and rr:propertyObjectMap are the fundamental properties

… for each row needs a subject

… all the triples for a particular row should be generated in the following way

… rr:propertyObjectMap is a pair propNameURI and propValue

… rr:computedPropertyMap if propName is not a constant

… maybe with some conversion specifications

… rr:rowGraphPropertyMap specifies destination graph for all triples generated for a particular row

… rr:foreignKeyMap needs a key for the property name

… and what is the join condition

… Section 3.4

… rr:ForeignKey that specifies the key

s/rr:key

/rr:key/

… going to examples section

… mapping specification for tables in Appendix B.2

… specification for the DEPT table

… this examples shows that we get 4 triples from one row

… in the rr:propertyObjectMap properties

… first is a literal of type positive integer

mhausenblas: why is the inderection with deptno and deptid?

Souri: value of bnode will be dept+100

… for example

mhausenblas: but why deptId and deptNo?

Souri: deptno is actually a number and you need a uri or bnode to use as identifier

… the generated triple for the prop will dept100

mhausenblas: what would happen if there was no deptNo in the table (?)

<juansequeda> deptId = "dept"+ deptno

<betehess> you mean that || is the concatenation operator over strings?

<juansequeda> deptID = "dept"+100 = dept100

<juansequeda> hence, deptID != deptno

<Seema> if there was no deptNo in the table..we would probably use some other column of the underlying table to generate the URI

MacTed: one of the problems is that the same string is used in multiple places

Souri: we can change dept to department in the string expression

… the point is that we are trying to generate an id for the row

… to be identified with the subject

mhausenblas: this is just because the subject is a bnode

… why should we have instructions for the label of the bnode?

… and not auto generated

<Zakim> mhausenblas, you wanted to ask about structure

Souri: it's a matter of convinience for readability of bnodes

mhausenblas: regarding structure of document, i'm missing something to group the classes and props

RRSAgent: draft minutes

cygri: so far the focus was on writing what it means

… in the next few days this will be a focus of ours

… better structure than just a listing of properties is better

hhalpin: all the action occurs inside literal values ?

Souri: no, not necessarily

hhalpin: this is close to d2rq?

<hhalpin> that's SQL

pathayes: looking at example B2.2 the quoted text is SQL but for RDF this is free text

… should I think of it as a string or something that is parsable?

… if so should the syntax be part of the spec/

Souri: no the syntax is not part of the spec

pathayes: we should mention that it must be valid SQL

Souri: yes

pathayes: the difference to D2Rq?

cygri: here you can write a complete SQL query

… while in D2RQ you have to split the query in parts and distribuite it for the specific part

Souri: in D2R we have a vocabulary

… here we can write the complete string

mhausenblas: next steps

… next try to sort out CVS issues

…. please send edits to the list so we can publish by next week?

cygri: we need some input to decide if we have achieved FPWD status

mhausenblas: I would be fine with the updates you do during this week

cygri: can you write a short email to the list with some points on what to be covered?

ericP: right now all the examples are in the appendix

<mhausenblas> ACTION: Hausenblas to sum up requirements for FPWD and send to editors and list [recorded in http://www.w3.org/2010/10/05-rdb2rdf-minutes.html#action01]

<trackbot> Created ACTION-74 - Sum up requirements for FPWD and send to editors and list [on Michael Hausenblas - due 2010-10-12].

… everyone in the WG should look at the syntax and make sure it is readble

juansequeda: for the logical table property why not create a SQL expression data-type

mhausenblas: i'll send the requirement right away, you can work until Friday and then I can work on the pub rules

… then if the group is ok with that we can have a formal resolution next week

<mhausenblas> [adjourned]

RRSAgent: draft minutes

<hhalpin> export CVSROOT=seema@cvs.w3.org:/w3ccvs