W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > April to June 2011

Re: while we are rechartering.... (csv)

From: Steve Harris <steve.harris@garlik.com>
Date: Wed, 1 Jun 2011 15:08:04 +0100
Cc: Bijan Parsia <bparsia@cs.man.ac.uk>, SPARQL Working Group <public-rdf-dawg@w3.org>
Message-Id: <82EFF12D-7A86-4890-9AAD-A33F7DE48CEF@garlik.com>
To: Sandro Hawke <sandro@w3.org>
On 2011-06-01, at 14:56, Sandro Hawke wrote:

> On Wed, 2011-06-01 at 14:38 +0100, Bijan Parsia wrote:
>> On 1 Jun 2011, at 14:32, Sandro Hawke wrote:
>> 
>>> TimBL suggests we also add a SPARQL CSV Results Format.
>>> (comma-separated-values, RFC 4180).
>>> 
>>> I think this is a good idea, probably worthwhile.  Particularly with
>>> government data, I see CSV is surprisingly popular.
>>> 
>>> (The use case is: some people want to publish CSV data, because they
>>> have a consumer community who wants it.  Rather than write any code, if
>>> we do this, they can just construct a SPARQL query to provide the data,
>>> give people a long URL (including the query) for getting that data as
>>> CSV.  Or they could hide the long URL behind a short one, via a proxy of
>>> some sort.  I find this use case rather compelling.)
>>> 
>>> Reactions?   like it, don't care, formally object, ...?
>> 
>> 
>> It's hugely useful. Lots of things consume CSV best (e.g., lots of google stuff) so on many of our LOD projects we end up hacking some XSLT or other processor in the middle.
>> 
>> It would really have to be "ready to go" as CSV rather than a csv encoding of the e.g., xml formats, or it's quite pointless.
> 
> Agreed.   I think the CSV for first example (SPARQL Query11/1.2.3)
> should be:
> 
>        x,y,z
>        Alice,http:/example/a,

We use TSV SPARQL results very extensively in Garlik, we actually emit turtle constants between tabs, rather than "ready to go" strings though. I think a fair number of 4store users use the format too, it's extremely fast to parse in perl, python, php etc.

e.g.

?x	?y	?z
"Alice"	<http://example/a>	_:b12

I'm open to being persuaded that the Turtle-ness should be dropped, though it has significant advantages when you want to push results back into SPARQL queries. I'd definitely be in favour of Turtle/string format being selectable, default off.

- Steve

> The key thing is that the results format should not indicate datatypes
> or node-IRI-vs-literal distinction -- if people want that information,
> they should SELECT it and put it in another column.
> 
>    -- Sandro

-- 
Steve Harris, CTO, Garlik Limited
1-3 Halford Road, Richmond, TW10 6AW, UK
+44 20 8439 8203  http://www.garlik.com/
Registered in England and Wales 535 7233 VAT # 849 0517 11
Registered office: Thames House, Portsmouth Road, Esher, Surrey, KT10 9AD
Received on Wednesday, 1 June 2011 14:08:34 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:46 GMT