New JSON query structure

[tfrancart]

• Objectives:
  • follow Traqula algebra structure as it will replace sparqljs
  • specify the query structure as a strict extension of the Traqula SPARQL algebra
  • enhanced compatibility with LLMs:
    • ability to explicitely embed children criterias in the object part of the criteria (so we understand they are tied to the object)
    • avoid repetition of the common subject in a serie of AND criterias (introduce a bgpSameSubject kind of pattern)
    • clearly associate the variable + type in a single entity
    • clearly associate the label to the variable (introduce a LabelledVariable kind of pattern)
    • refactor the "order" part like a standard order clause. When loading a query in the UI, if the order clause is not on a single variable which is the first one, ignore it
• new features:
  • ability to express prefixes in the query
  • clear disctinction between values in the query vs. filters. Any fixed value, be it an IRI or a Literal, would go in a values clause and not a filters clause

Test query

Proposed example structure:

{
  // all this to mimic Traqula
  "type": "query",
  "subType": "select",
  "distinct": true,
  "variables": [
    {
      "type" : "term",
      "subType": "variable",
      "value": "Artwork_1"
    }
  ],

  // "order": null,
  // change order to mimic sparqljs, but in our case this can be only the first variable in the query
  "solutionModifiers": {
    "order": {
      "type": "solutionModifier",
      "subType": "order",
      "orderDefs": [
        {
          "expression": {
            "type": "term",
            "subType": "variable",
            "value": "Artwork_1"
          },
          // si on a besoin de mettre un DESC
          // "descending": true
        }
      ]
    },
    "limitOffset": {
      "type": "solutionModifier",
      "subType": "limitOffset",
      "limit": 1000
    }
  },

  // change "branches" to "where"
  // and this is a single object, not an array. The array is bgpSameSubject
  //  "branches": [
  "where": {
    "type" : "pattern",
    "subType" : "bgpSameSubject",
    "subject": {
      // new type variable with an rdf:type
      "type": "term",
      "subType": "variable",
      "value": "Artwork_1",
      "rdfType": "https://data.mydomain.com/ontologies/sparnatural-config/Artwork"      
    },
    // list of predicate object pairs associated to the same subject declared above
    "predicateObjectPairs": [
      {
        "type": "predicateObjectPair",
        "predicate": {
          "type": "term",
          "subType": "namedNode",
          "value": "https://data.mydomain.com/ontologies/sparnatural-config/Artwork_displayedAt"
        },
        "object" : {
          "type": "objectCriteria",
          "variable": {
            "type": "term",
            "subType": "variable",
            "value": "Museum_2",
            "rdfType": "https://data.mydomain.com/ontologies/sparnatural-config/Museum"  
          },
          // either "values", either "filters", either "predicateObjectPairs"
          // here a predicateObjectPair means we have a WHERE associated to the variable declared above
          "predicateObjectPairs": [
            {
              "type": "predicateObjectPair",
              // here : if we need to switch optional / notExists this will be a "subType"
              // "subType": "optional",
              // "subType": "notExists",
              "predicate": {
                "type": "term",
                "subType": "namedNode",
                "value": "https://data.mydomain.com/ontologies/sparnatural-config/Museum_country"
              },
              "object" : {
                "termType": "objectCriteria",
                "variable": {
                  "type": "term",
                  "subType": "variable",
                  "value": "Country_4",
                  "rdfType": "https://data.mydomain.com/ontologies/sparnatural-config/Country"
                },
                // we would always have "values", and if we have a single value, we set
                // it as the object value instead of a VALUES clause
                // fixed values set by "values" clause instead of "filters"
                "values": [
                  {
                    // something closer to Traqula than to rdfjs here
                    "type": "term",
                    "subType" : "namedNode",
                    "value": "http://dbpedia.org/resource/France",
                    "label": "France"
                  },
                  {
                    "type": "term",
                    "subType" : "labelledIri",
                    "value": "http://dbpedia.org/resource/Italy",
                    "label": "Italy"
                  }
                ]                     
              }                     
            }
          ]
        }
      }
      {
        "type": "predicateObjectPair",
        "predicate": {
          "type": "term",
          "subType": "namedNode",
          "value": "https://data.mydomain.com/ontologies/sparnatural-config/Artwork_author"
        },
        "object" : {
          "type": "objectCriteria",
          "variable": {
            "type": "term",
            "subType": "variable",
            "value": "Person_6",
            "rdfType": "https://data.mydomain.com/ontologies/sparnatural-config/Person"
          },
          "values": [
            {
              "type": "term",
              "subType" : "namedNode",
              "value": "http://dbpedia.org/resource/Leonardo_da_Vinci",
              "label": "Leonardo da Vinci"
            }
          ]
        }
      },
      {
        "type": "predicateObjectPair",
        "predicate": {
          "type": "term",
          "subType": "namedNode",
          "value": "https://data.mydomain.com/ontologies/sparnatural-config/Artwork_creationYear"
        },
        "object" : {
          "termType": "ObjectCriteria",
          "variable": {
            "type": "term",
            "subType": "variable",
            "value": "Date_8",
            "rdfType": "https://data.mydomain.com/ontologies/sparnatural-config/Date"
          },
          // here a LabelledFilter, that is a custom filter with a label
          // note : fixed literal values, like for a Boolean widget, would go in a "values" clause
          "filters": [
            {
              "type": "labelledFilter",
              "label": "between 1800 and 1901",
              "filter": {
                // here we would have different kind of filters depending on the widget
                "type": "numberFilter",
                "min": 1800,
                "max": 1901
              }
            }
          ]
        }
      }
    ]
  }
}

[tfrancart]

fyi @BouzelatTarek

[tfrancart]

Traqula Typescript Algebra definitions: https://github.com/comunica/traqula/blob/ac278a201bf487889b063168c056b5495b9ed6b5/packages/algebra-transformations-1-1/lib/algebra.ts
Traqula AST definitions: https://github.com/comunica/traqula/blob/ac278a201bf487889b063168c056b5495b9ed6b5/packages/rules-sparql-1-1/lib/Sparql11types.ts

[tfrancart]

faire un exemple avec un literal en values

Avec un datatype:

    "values": [
      {
         "type": "term",
         "subType": "literal",
         "value": "true",
         "langOrIri": {
            "type": "term",
            "subType": "namedNode",
            "value": "http://www.w3.org/2001/XMLSchema#boolean"
         }
    ]  

Avec une langue

    "values": [
      {
         "type": "term",
         "subType": "literal",
         "value": "valeur",
         "langOrIri": "fr"
    ]  

donner les listes et des exemples des filters précis

Number:

{
                "type": "DateFilter",
                "min": 1800,
                "max": 1901
              }

Search:

{
                "type": "SearchFilter",
                "search": "xxxxxxx"
              }

Date:

{
                "type": "NumberFilter",
                "begin": "1800-01-01",
                "end": "1901-09-10"
              }

Map:

{
                "type": "MapFilter",
                "coordType": 'Polygon' | 'Rectangle';
                "coordinates": LatLng[][];
              }

structures par widget

Widget	structure critère
List	values or predicateObjectPairs
Autocomplete	values or predicateObjectPairs
Tree	values or predicateObjectPairs
SearchRegex	filter or predicateObjectPairs
Date (date or year)	filter or predicateObjectPairs
Number	filter or predicateObjectPairs
Boolean	values or predicateObjectPairs
Map	filter or predicateObjectPairs
NoWidget	predicateObjectPairs