Re: On the inevitability of SPARQL/SPIN for SHAQL

Jose,

> On 3 Mar 2015, at 04:59, Jose Emilio Labra Gayo <jelabra@gmail.com> wrote:
> 
> I was talking about the same subset of XPath functions that is used in the Filter Expressions in SPARQL...as Richard said, you can call it Sparql expressions…

What you say is confusing. SPARQL expressions are not a subset of XPath. It’s a very different language.

Do you propose that the WG design a new expression language for operating over RDF nodes that is an XPath subset?

Or do you propose support for SPARQL expressions as-is?

> That subset has a simple semantics that we can leverage on and does not have the complexity of the full SPARQL. 

I’m not sure that I would agree with this statement. A lot of the complexity of SPARQL is in the expression language and function library.

This is difficult to quantify, but in my printed copy of the SPARQL spec, the chapters describing the various language constructs cover 46 pages. The single chapter on SPARQL expressions is the longest, covering 20 pages. And most of the functions are not defined in those 20 pages, but are only defined by reference to the XPath spec. In O’Reilly’s book “Learning SPARQL”, authored by my colleague Bob DuCharme (which by the way I highly recommend to you), the language feature chapters are 130 pages, and the expression language covers 50.

For comparison, the full XPath Functions and Operators spec is 152 printed pages.

I agree however that we should leverage prior work on specifications and implementations as much as possible.

Richard



> 
> Anyway, I had proposed it as one of two possible solutions to handle complex constraints. The other possibility is just to use the extensibility mechanism which would allow any SPARQL query.
> 
> I now believe you are proposing designing a new language that, like SPARQL, would operate on RDF graphs to match graph patterns and bind variables and, like SPARQL, would include functions and operators (e.g., addition) borrowing definitions of some of these from XQuery/XPath. 
> 
> Why should this group take on such undertaking instead of reusing already existing language produced by W3C?
> 
> Because "SPARQL queries cannot easily be inspected and understood, either by human beings or by machines, to uncover the constraints that are to be respected". [1]
> 
> Best regards, Jose Labra
> 
> [1] Conclusions of RDF Validation Workshop. http://www.w3.org/2012/12/rdf-val/report
> 
> On Mar 1, 2015, at 11:41 PM, Jose Emilio Labra Gayo <jelabra@gmail.com> wrote:
> 
>> On Mon, Mar 2, 2015 at 4:01 AM, Dean Allemang <dallemang@workingontologist.com> wrote:
>> ARQ (and hence, SPIN) use a lot of the xpath functions in various sensible place (filter, bind).  xpath certainly seems like a good place to go for your built-in function vocabulary.  I like to idea of being able to extend this (as TopBraid does), but I don't know if there is a way to standardize the extensions. 
>> 
>> If that is what you are proposing, then I think I find myself in violent agreement; we should certainly use the function set from XPath for Shapes - no sense in re-inventing that, and we certainly need a function vocabulary. 
>> 
>> Yes, that is more or less what I was proposing. 
>> 
>> My proposal was to have a construct in the language to bind variables to the subjects/objects that are matched and another construct that allows simple expressions using those variables. 
>> 
>> But my proposal was to limit those expressions to use the same expressions that can be used in the "FILTER" expression of SPARQL queries instead of any kind of SPARQL query, that's what I call "a controlled way"
>> 
>> Best regards, Jose Labra
>> 
>> 
>> 
>> 
>> 
>> Dean
>> 
>> 
>> On Sun, Mar 1, 2015 at 6:58 PM, Irene Polikoff <irene@topquadrant.com> wrote:
>> Jose,
>> 
>> I do not understand what you are proposing - how exactly you propose to use XPath, what do you mean by 'controlled way' and so on.
>>  
>> I feel these ideas are too vague and confusing to have an effective conversation about.
>> 
>> Irene
>> 
>> On Mar 1, 2015, at 3:57 PM, Jose Emilio Labra Gayo <jelabra@gmail.com> wrote:
>> 
>>> On Sun, Mar 1, 2015 at 5:03 PM, Dean Allemang <dallemang@workingontologist.com> wrote:
>>> While I am a big fan of XPath (I use it all the time), I have to agree with Irene here - treating RDF as if it were XML seems like a bad idea. 
>>> 
>>> As I said in my previous email. My proposal was to use the same subset of XPath that SPARQL is using in the FILTER expressions so the Shacl language can define constraints that involve arithmetic and string operations in an easy and controlled way.
>>> 
>>> I can't help but think that there is a perception thing going on.  The resistance to committing to SPARQL as a constraint language in favor of, e.g., XPath,
>>> 
>>> I think there is a misconception in that phrase. In fact, if we use SPARQL we will also use the same subset of XPATH. My proposal was to use only that subset as predefined builtin expressions...
>>> 
>>> seems to stem from a feeling that SPARQL is a low-level, implementation-specific language whereas some alternative is higher-level.  
>>> 
>>> I view SPARQL as being at a much higher level than, e.g., XPath (just as I view RDF as being at a higher level than XML).  In RDF, we talk about how our resources relate to one another - not how a document (or a table, or a spreadsheet, or etc.) happens to structure that relationship.   XPath is a high-level language for navigating a very particular structure; if you change the structure, your XPath is at risk (and many of the features of XPath are there to help you make an XPath expression immune to certain changes in structure).   SPARQL is a language for navigating data graphs, regardless of how they are represented in a document.  
>>> 
>>> 
>>> I think this is the source of a lot of the "XML Confusion" that Irene mentions in her message.  Those of us who are RDF fans can't help but be puzzled by any language proposal that, in our view, makes lower-level commitments in the shapes language.  
>>> 
>>> I suspect that there are others for whom SPARQL seems like the low-level language.  I can't represent this viewpoint as well since I don't myself hold it.  But if this is the view one is coming from, then committing to SPARQL would seem like a low-level commitment.  
>>> 
>>> and I think we are agreed that we expect Shapes to develop a high-level language.  We just don't agree on where the levels are. 
>>> 
>>> Yes, I also want Shapes to be a high level language...that's probably the main point that I am trying to defend. If we want to have a high level language, we should not embed SPARQL inside it without control. 
>>> 
>>> Best regards, Jose Labra 
>>> 
>>> 
>>> 
>>> 
>>> Dean
>>> 
>>> 
>>> On Sun, Mar 1, 2015 at 6:19 AM, Irene Polikoff <irene@topquadrant.com> wrote:
>>> Personally, I feel that mixing paradigms results in design that is inelegant and problematic on many levels. This is RDF. Why bring XPath in when there are equally good and better approaches within RDF stack?
>>> 
>>> Besides RDF is still recovering from XML baggage and misunderstandings that resulted from people confusing it with its XML serialization. 
>>> 
>>> Irene
>>> 
>>> Sent from my iPhone
>>> 
>>> On Mar 1, 2015, at 4:45 AM, Jose Emilio Labra Gayo <jelabra@gmail.com> wrote:
>>> 
>>>> On Sun, Mar 1, 2015 at 9:56 AM, Holger Knublauch <holger@topquadrant.com> wrote:
>>>> 
>>>> On 3/1/15 5:24 PM, Jose Emilio Labra Gayo wrote:
>>>> 2.- To allow the core language to have XPATH like functionality and variables. It can be something similar to the expressions that appear in the FILTER clauses in SPARQL. As an example using the compact syntax we could say:
>>>> 
>>>> <RectangleShape> { :weidth ?w, :height ?h, :area ?a, FILTER (?w * ?h = ?a) }
>>>> 
>>>> This looks like it would simply reinvent a new SPARQL, only that no existing tool would support it yet.
>>>> 
>>>> No, it is leveraging on XPath which has lots of implementations and tools. Maybe, the syntax can be "CONSTRAINT" instead of "FILTER" to clarifiy that it is not SPARQL. 
>>>> 
>>>> <RectangleShape> { :weidth ?w, :height ?h, :area ?a, CONSTRAINT (?w * ?h = ?a) }
>>>> 
>>>> The only neede feature is to associate variables with the objects that are being matched and to have a "CONSTRAINT <XPath-Expr>" that evaluates to a boolean. 
>>>> 
>>>> But generating human-readable error messages can be a post-process operation. Embedding that functionality in SPARQL you are preventing any implementation that is not based in SPARQL.
>>>> 
>>>> Why not?
>>>> 
>>>> Because  you are embedding SPARQL in the generation of human-readable messages.
>>>> 
>>>> If someone wants to use another language, then this would also have a mechanism to create strings. JavaScript certainly has.
>>>> 
>>>> And in that way, the shapes generating human-readable messages in SPARQL are not compatible with the shapes generating human-readable messages in Javascript. 
>>>>  
>>>> And even if not, the human-readable messages are purely optional anyway, but preventing something that is already solved by SPARQL doesn't sound like a good idea.
>>>> 
>>>> Because we are using SPARQL for something that is not needed. 
>>>> 
>>>> And I emphasize, I am not against SPARQL, I am against embedding SPARQL in an uncontrolled way what is supposed to be a high-level language.
>>>> 
>>>> Best regards, Labra
>>>> 
>>>>  
>>>> 
>>>> Holger
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> -- 
>>>> -- Jose Labra
>>>> 
>>> 
>>> 
>>> 
>>> 
>>> -- 
>>> -- Jose Labra
>>> 
>> 
>> 
>> 
>> 
>> -- 
>> -- Jose Labra
>> 
> 
> 
> 
> -- 
> -- Jose Labra
>

Received on Tuesday, 3 March 2015 11:25:12 UTC