Re: Templating: a proposed way forward

Hi all, 

Still getting up to speed with the working group.... The work here is much
more theoretical than I'm used to, and I'm trying to figure out what I can
contribute. I am much stronger in applied research -- trying to make the
things people like you come up with work in a research library -- instead
of coming up with those things myself. So bear with me.

I have been using XSLT as an analogy to help me understand the work going
on here. For instance, an XML file can be transformed into, say, (X)HTML
by applying a stylesheet. Likewise, a CSV file can be transformed into,
say, XML, which is what we're trying to accomplish. Is that a fair
analogy? If so, then it follows that what's missing from the CSV-to-XML
transformation is the "stylesheet".

So what is the problem this working group is trying to solve? Is is to
make stylesheets (templates) for transforming CSVs into XML/RDF/JSON in
order to demonstrate the conversions? Is it to specify mappings and make
recommendations for how conversions should be made in general? Or is it to
create an XSLT-like language for doing transformations on CSVs?

In the case of the latter, this seems like a huge undertaking. XSLT is a
Turing-complete programming language! Is the concern that trying to define
a basic template language would quickly become complicated for the reasons
Jeremy pointed out on the call, stated below? That's my concern, but maybe
I'm overlooking something obvious.

By the way, I've just bought my ticket to TPAC. I'm looking forward to
meeting you in person.

Cheers, 
Bill



On 9/22/14, 4:27 AM, "Ivan Herman" <ivan@w3.org> wrote:

>
>On 22 Sep 2014, at 10:30 , Jeni Tennison <jeni@jenitennison.com> wrote:
>
>> Hi all,
>> 
>> Thanks to all those who managed to attend the telcon on Wednesday.
>>Those of you who didn¹t, please read the minutes at
>>http://www.w3.org/2014/09/17-csvw-minutes.html for background.
>> 
>> Based on our charter, we need to deliver a default mapping of CSV to
>>RDF/XML/JSON, informed by the basic metadata specified in [1]. These
>>conversions will need to address some particular user needs, which might
>>include:
>> 
>>   * specifying the names/URIs of properties/elements/attributes
>>   * specifying the datatypes of values in the target format
>>   * specifying how values map to URIs where appropriate
>>   * handling list values
>>   * handling default values when values are missing
>>   * converting sets of CSV files into a single output
>> 
>> There is a limit to what can be easily articulated through properties
>>in a metadata document however, without it becoming extremely unwieldy.
>> 
>> Jeremy¹s analysis of the use cases demonstrate that we have very few
>>use cases that include a specification of the converted format, and
>>those that do are relatively simple. I think it was also Jeremy who
>>pointed out on the call that the places where conversions are usually
>>hard are when targeting pre-defined formats such as Data Cube or vCard,
>>where there need to be more complicated mappings between the data you
>>have and the format you want to create.
>> 
>> On the call, Stasinos suggested:
>> 
>>   * we define how to point out to templates for conversion from
>>metadata files
>>   * we support implementation defined (extension) template languages
>>   * we define a basic template language and conformance level around
>>that
>
>To be clear: do we (should we) consider this as yet another option
>compared to what was discussed on the call? Of course, the devil is in
>the details: what does 'basic' mean? Does it mean that the language does
>not cover all the issues listed above?
>
>> 
>> Andy suggested the the work to define a basic template language could
>>be done within a Community Group rather than within this Working Group,
>>depending on the interest and time commitment that the Working Group can
>>make.
>> 
>> I want to reiterate one particular point that was made during the
>>meeting: the Working Group as a whole can only do what the members of
>>the Working Group do. Regardless of whether we pursue the specification
>>of a templating language, we still have need for volunteers to edit the
>>specs for:
>> 
>>   * default conversion to RDF
>>   * default conversion to XML
>>   * default conversion to JSON
>> 
>> Perhaps the biggest question we have about a templating language is
>>actually about the Working Group being able to commit the time that it
>>would take to specify it, including all the tests that would be required
>>and the need for implementations. Having someone volunteer to edit the
>>spec, someone volunteer to manage the test suite, and two people
>>volunteer to implement the spec (some of these volunteers may be the
>>same people), would make me at least feel a lot happier about taking
>>this on within the Working Group.
>> 
>> Anyone feel they can step up?
>> 
>> We will discuss this again this week. If we can¹t decide then we¹ll
>>defer a decision until the face-to-face at TPAC at the end of October.
>> 
>> Cheers,
>> 
>> Jeni
>> 
>> [1] http://w3c.github.io/csvw/metadata
>> --  
>> Jeni Tennison
>> http://www.jenitennison.com/
>> 
>
>
>----
>Ivan Herman, W3C 
>Digital Publishing Activity Lead
>Home: http://www.w3.org/People/Ivan/
>mobile: +31-641044153
>GPG: 0x343F1A3D
>WebID: http://www.ivan-herman.net/foaf#me
>
>
>
>
>