[ckan-discuss] CKAN CSV alert

[koumenides c.l. (clk1v07)]

for dates, I basically used a set of rules in my conversion to realign the month and day intervals to be consistent to the ISO8601 format in all the dates. For dates that only the year was available I used 01-01 for months and days. I can't remember how thorough my scripts were (probably > 80%), but I was very satisfied and used them for a visualization in one of my papers.

I would also advice for URIs for tags, avoid any bnodes (i.e. b36188902000082 in http://world.ckan.net/data.gov.uk/data/emissions-of-ammonia), and correct use of dcat:distribution (ideally should link to a dcat:Download and not the csv file). Assuming that you do plan to include proper uris for agencies, authors, departments, and not stay with bnodes.

p.s. it is not very clear in the dcat spec if the property dcat:dataset has a CatalogRecord for domain, but it does make sense to use it that way.

Christos
________________________________________
From: William Waites [william.waites at okfn.org]
Sent: 09 June 2010 12:55
To: Rufus Pollock
Cc: koumenides c.l. (clk1v07); CKAN discuss; ms8 at ecs.soton.ac.uk
Subject: Re: [ckan-discuss] CKAN CSV alert

On Wed, Jun 09, 2010 at 11:19:18AM +0100, Rufus Pollock wrote:
>
> Any patches for date cleaning/validation etc would be gratefully
> received I imagine by us and the data.gov.uk rdf team :)
>

quite. as far as i am aware the main problwms are invalid dates and
invalid resource urls. this makes turning the data into valid rdf very
hard.

cheers,
-w