Re: [xml-dev] RE: Auto-generating XML documents from Schematron (was:N-Q

[Rick Jelliffe <rjelliffe@allette.com.au>]

Is the issue really 'Can Alloy generate models etc that Schematron cannot?' Or is it 'can a Schematron/XPath schema be translated into an Alloy model?'

Seeing that Michael made an XPath 1.0 data model to Alloy converter, it seems likely to me that the conversion is entirely possible, to some degree of completeness. There are also papers detailing the translation of XPath into predicate logic IIRC that would provide more info. 

Then the issue would be whether the particular schema is closed and complete enough to generate documents, I expect.

Regards

Rick

On 6 Apr 2017 00:58, "Costello, Roger L." <costello@mitre.org> wrote:

Oops! My drawing of the chessboard had an error. Here’s the correct version:

 

 

 

From: Costello, Roger L.
Sent: Wednesday, April 5, 2017 10:49 AM
To: xml-dev@lists.xml.org
Subject: Auto-generating XML documents from Schematron (was: N-Queens problem)

 

Hi Folks,

As you know, Schematron can be used to express powerful data relationships. We saw an example of this with the N-Queens problem.

The Schematron document for the N-Queens problem can be used to validate XML instance documents, to check that they contain a valid solution to the N-Queens problem.

Great!

But, but, but, …

Finding a valid solution to the N-Queens problem can be difficult, particularly for large chessboards.

It would be useful if we could auto-generate valid solutions directly from the Schematron document.

I am not aware of any tool that can auto-generate XML instances from Schematron.

With Alloy, you can express the same powerful data relationships that you can express with Schematron. Plus, the Alloy Analyzer can auto-generate solutions. In fact, the Analyzer can auto-generate a comprehensive set of solutions.

For example, at the bottom of this message is an Alloy model of the N-Queens problem. I instructed the Alloy Analyzer to generate solutions for a 5 x 5 chessboard. The Analyzer generated a complete set of solutions. Here are two of the solutions that it auto-generated:

Fantastic!

/Roger

Alloy Model (thanks Loïc Gammaitoni)

sig Queen {
    column: Column,
    row: Row
}

fact Queens_Not_On_Same_Row_Or_Column {
    no disj q, q': Queen | q.column = q'.column
    no disj q, q': Queen | q.row = q'.row
}

fact Queens_Not_On_Diagonal {
    no disj q, q': Queen |
        some a,b: Int {
            add[q.column.id, a] = q'.column.id
            add[q.row.id, b] = q'.row.id
            a=b or a=sub[0,b]
        }
}

…