CSV on the Web Working Group Teleconference

01 Oct 2014


See also: IRC log


Yakov Shafranovich (yakovsh), Ivan Herman (Ivan), Dan Brickley (danbri), Phil Archer (phila), Andy Seaborne (AndyS), Jeremy Tandy (jtandy)
Jenit, William
Dan Brickley
Andy Seaborne


<trackbot> Date: 01 October 2014

Review Ivan's progress on RDF and JSON mappings

<jtandy> Apologies - but I did read the minutes

<danbri_> see thread: http://lists.w3.org/Archives/Public/public-csv-wg/2014Sep/0084.html

danbri_: what is the status? what help do you (Ivan) want?

<jtandy> I agree that starting with a minimal approach is a good thing

ivan: minimal start the add things.
... the new doc is the minimal mapping from metadata, no templating.
... work done in the spring. Output is RDF triples, not syntax.
... proper way to do it for RDF.

<danbri_> (sorry AndyS, typing makes ivan hard to hear!)

ivan: Then with minimal changes, same for JSON. Believe it is 80% JSON-LD.

<danbri_> JSON-LD is here, http://w3c.github.io/csvw/csv2rdf/index-json.html

ivan: for each row, series of literal-values properties.
... mint URI for keys.
... metadata does not describe objects as URIs.
... similar to "direct mapping" of RDB2RDF.
... JSON same style, one row => one object
... datatypes used from metadata.
... row number added to each row and some top-level DC (schema.org?) as global triples. Pass through.
... not handled : local template for a column (object) and ditto predicates.
... issues in the document
... one issue ignored, can have no column names (a lot of metadata is optional).
... more for the metadata work.
... one use of output is feed into a later sparql construct named in metadata.
... for JSON, unclear as to best tool to name here.
... for XMl, XSLT (XQuery?)
... personally, not decided as to whether this is the approach to follow.

jtandy: Made sense. Will work on a real case for the F2F.
... compound keys not covered.
... assigning identifiers from other columns
... URI templates

ivan: primary key of several cols is in the dc as the concat with "-"
... if you want a specific form for row or col, assuming a template.

jtandy: if we can stick to simple URIs (??), then using URI template (RFC).

<Zakim> danbri_, you wanted to propose "No mapping is defined, in the absence of column names within file or external metadata"

ivan: just add issues to the doc.

danbri_: suggest - if no cols, then no mapping.

ivan: suggestion is "col_1", "col_2", ...

danbri_: project tables from tables
... needs col names to be given.

<yakovsh> http://tc39wiki.calculist.org/es6/template-strings/

yakovsh: javascript will have a templating language built-in.
... early next year

ivan: But for python, CSV->JSON->Python->

<Zakim> jtandy, you wanted to ask if there is a minimum metadata requirement ... column headings _either_ in the csv _or_ in the metadata?

jtandy: Starting from min approach is fine, maybe have templating as second pass.
... state a min as some source of column names (several choices).

<danbri_> ThrowUhUhException()

phila: If 80% cover from direct of the UC, then good.
... would it be possible to say, a preprocess step (c.f. second pass)?
... want to see URI templating
... should check the UC against the proposed solution.

ivan: How to spec that (processing chain) properly?

danbri_: (1) processing chain - lots of different possibilities - not for us to dictate this set. We need to write it down.
... record the chain in the metadata
... (2) ?
... Ivan, what do you us to do with this in the next 3 weeks and at F2F?

ivan: Jeremy's plan (worked UC) is very useful
... an implementation would be good.
... not ready to do the same for XML.
... don't want to be the champion for this. Not the only editor. Can contribute.
... good to check the JSON structure is JSON-LD == RDF if @context added.

<phila> Gregg is indeed registered for the f2f at TPAC

<danbri_> ACTION: danbri take a look at json-ld assuming very naive @context file [recorded in http://www.w3.org/2014/10/01-csvw-minutes.html#action01]

<trackbot> Created ACTION-31 - Take a look at json-ld assuming very naive @context file [on Dan Brickley - due 2014-10-08].

ivan: @context is complicated
... whether it can handle things is unclear to me.

jtandy: I offer to contribute to direct mapping.

<phila> Erik Mannens, Axel Polleres, William Ingram

phila: people at F2F above.
... maybe some capacity here.

danbri_: anything else on Ivan's doc?

ivan: avoided @graph usage needed for a graph, not tree.

<ivan> { ... "rows" : [ {row1},{row2} ] }

(continued) Which metadata vocabulary/ies should we be using? (DC, schema.org, ...)

danbri_: Not sure what we can do this week on this.
... W3C has a problem with refs to non-standards orgs.
... for mailing list and F2F.

<jtandy> JeniT noted that the data on the web best practice group is recommending metadata vocabs; we need to be aligned.

ivan: Still an issue because normative refs means mandatory but some people use Dublin Core. And other vocabs.

Surely this is the point of a standard?

danbri_: avoid disruption.
... will go to email.

<Zakim> phila, you wanted to agree with everyone else and...

phila: agree and wonder whether data shapes work helps here.

<danbri_> (AndyS - yeah, the question is what level to standardize at... e.g. triples, use cases, specific vocabs, or [per Phil's point] specific graph patterns)

phila: DCAT? A W3C std.

<danbri_> http://www.w3.org/TR/vocab-dcat/

danbri: how about DCAT + owl:equivalent?

It's too rigid.

F2F planning

ivan: agenda needed

<danbri_> AndyS: issue-centric discussions often more affective

<danbri_> esp when non-telecon-attendees are joining

<jtandy> agree with AndyS to focus on specific issues ... with outcome as being "we decided X Y Z"

Summary of Action Items

[NEW] ACTION: danbri take a look at json-ld assuming very naive @context file [recorded in http://www.w3.org/2014/10/01-csvw-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2014-10-01 13:16:29 $