How to use Smart Data Models in your projects

This section aims to provide few simple guidelines for the adoption of Smart Data Models. Readers interested into modifying or creating new data models should refer to Data models guidelines. This guide is not exhaustive and does not aim to cover the specifics of each model, rather it provides general usage tips valid for most of the existing models and for expected models in the future.

Data Models have been defined in relation to the FIWARE reference data context model (NGSI v2). While this does not imply that they cannot be used outside of the NGSI context model, it does however indicate that some of design principles have been driven by that. This also implicitly means that Data Models will follow the evolution of FIWARE platform, NGSI-LD mappings are being added to all existing Data Models.

General principles

The general principle guiding your adoption should be: keep it simple.

  • Use only the parts of the Data Model your application really need. This will help reducing the amount of information exchanged to the one you need and not more than that. This may apply also just to a part of your application. I.e. while you use in your Context Broker the whole Data Model, to other services you may dispatch only the needed part.
  • Use metadata only in case of need, and follow the assumptions provided in the Data Model. For example, in case your application is leveraging the default unit defined in a Data Model (e.g. meters), avoid attaching metadata to annotate the unit. In case your application is using a different unit, do annotate it, it will simplify data exchange with other systems in case of need.
  • Stick to the existing Data Models. Data Models are meant to be an effort to facilitate interoperability within the community. If you change the meaning or the data type attached to an attribute, this will break interoperability. Of course it may be the case that the model is not covering your scenario, or it contains errors. Open a ticket in the repository, and let the community discuss on the issue! Contribute to the evolution of the model! In doing so please keep in mind the guidelines to create new data models.
  • While not generally advised, you can add additional attributes in your specific application for the purpose of your application. Be aware that of course this means that such attributes may not be correctly used by other applications leveraging the same Data Model. We recommend, if the attribute you want to add is of general adoption and not specific to your case, to propose it to become officially part of the Data Model. You can open an issue or a pull request for that.
  • Data Models, in most of the cases, are NGSI v2 codification of existing data models. You can find the references in each Data Model specification, but for the sake of maintainability, on purpose we don't report the full specification. This means that to gather further details, you need to access the original source. For example, for the location, we use GeoJSON. You can find the exact semantics of GeoJSON schema in the related standard.
  • Be aware that JSON Schemas provided cover only the so called key-value representation of NGSI v2 context data.
  • Be aware that at the time being none of the official FIWARE Core Enablers enforce schema validation. Thus, in case you want to be sure your application data validates correctly against a given Data Model, this is up to you.

Going through a Data Model

Each Data Model is programmatically defined using a JSON Schema, as previously mentioned, the JSON schema covers only the so called key-value representation of NGSI v2 context data. Thus the JSON Schema does not cover the normalised representation of context data.

In the NGSI v2 normalised format each attribute (key) of the Data Model is represented by a JSON object with the following syntax:

  • The attribute value is specified by the value property, whose value may be any JSON datatype.
  • The attribute NGSI type is specified by the type property, whose value is a string containing the NGSI type.
  • The attribute metadata is specified by the metadata property. Its value is another JSON object which contains a property per metadata element defined (the name of the property is the name of the metadata element). Each metadata element, in turn, is represented by a JSON object containing the following properties:
    • value: Its value contains the metadata value, which may correspond to any JSON datatype.
    • type: Its value contains a string representation of the metadata NGSI type.
{
    "id": "entityId",
    "type": "entityType",
    "att1": {
        "value": "value1",
        "type": "Text",
        "metadata": {
            "metada1": {
                "value": "metavalue1"
            }
        }
    }
}

In the case of the key-value format (also known as Simplified Entity Representation in the NGSI v2 specification), values are directly mapped to the keys, and metadata are not covered:

{
    "id": "entityId",
    "type": "entityType",
    "att1": "value1"
}

Of course the normalised format contains richer information, but on the other side it is redundant and less efficient in term of transport.

To clarify the difference among the two representation here is a complete example from the AirQualityObserved Data Model:

  • Normalised format:
{
    "id": "Madrid-AmbientObserved-28079004-2016-03-15T11:00:00",
    "type": "AirQualityObserved",
    "dateObserved": {
        "value": "2016-03-15T11:00:00/2016-03-15T12:00:00"
    },
    "airQualityLevel": {
        "value": "moderate"
    },
    "CO": {
        "value": 500,
        "metadata": {
            "unitCode": {
                "value": "GP"
            }
        }
    },
    "temperature": {
        "value": 12.2
    },
    "NO": {
        "value": 45,
        "metadata": {
            "unitCode": {
                "value": "GQ"
            }
        }
    },
    "refPointOfInterest": {
        "type": "Relationship",
        "value": "28079004-Pza.deEspanya"
    },
    "windDirection": {
        "value": 186
    },
    "source": {
        "value": "http://datos.madrid.es"
    },
    "windSpeed": {
        "value": 0.64
    },
    "SO2": {
        "value": 11,
        "metadata": {
            "unitCode": {
                "value": "GQ"
            }
        }
    },
    "NOx": {
        "value": 139,
        "metadata": {
            "unitCode": {
                "value": "GQ"
            }
        }
    },
    "location": {
        "type": "geo:json",
        "value": {
            "type": "Point",
            "coordinates": [-3.712247222222222, 40.423852777777775]
        }
    },
    "airQualityIndex": {
        "value": 65
    },
    "address": {
        "type": "PostalAddress",
        "value": {
            "addressCountry": "ES",
            "addressLocality": "Madrid",
            "streetAddress": "Plaza de Espa\u00f1a"
        }
    },
    "reliability": {
        "value": 0.7
    },
    "relativeHumidity": {
        "value": 0.54
    },
    "precipitation": {
        "value": 0
    },
    "NO2": {
        "value": 69,
        "metadata": {
            "unitCode": {
                "value": "GQ"
            }
        }
    },
    "CO_Level": {
        "value": "moderate"
    }
}
  • key-value format:
{
    "id": "Madrid-AmbientObserved-28079004-2016-03-15T11:00:00",
    "type": "AirQualityObserved",
    "address": {
        "addressCountry": "ES",
        "addressLocality": "Madrid",
        "streetAddress": "Plaza de España"
    },
    "dateObserved": "2016-03-15T11:00:00/2016-03-15T12:00:00",
    "location": {
        "type": "Point",
        "coordinates": [-3.712247222222222, 40.423852777777775]
    },
    "source": "http://datos.madrid.es",
    "precipitation": 0,
    "relativeHumidity": 0.54,
    "temperature": 12.2,
    "windDirection": 186,
    "windSpeed": 0.64,
    "airQualityLevel": "moderate",
    "reliability": 0.9,
    "CO": 500,
    "NO": 45,
    "NO2": 69,
    "NOx": 139,
    "SO2": 11,
    "CO_Level": "good",
    "NO_Level": "moderate",
    "refPointOfInterest": "28079004-Pza. de España"
}

For each Data Model we provide a set of examples (in some cases also using the Normalised representation of NGSI v2), such examples have been validated against the respective JSON Schema and against a Orion Context Broker instance (to ensure that also the content encoding is compliant with the reference implementation).

As a consequence of the NGSI v2 specifications, all Data Models must include the following attributes:

  • id : a unique identified of the entity modelled.
  • type : The entity type, i.e. the type of Data Model, e.g. Alert.
  • dateModified : Last update timestamp of the entity.
  • dateCreated : The entity's creation timestamp.

dateCreated and dateModified are special entity attributes provided off-the-shelf by NGSI v2 implementations.

Most of the Smart Data Models adopt the GSMA common definitions, and so also include standard GSMA terminology such as:

  • owner: An array of URIs or pointers to NGSI entities representing the owner(s) of the entity.
  • source: A pointer (eventually an URI) to the service providing the data.
  • name: A mnemonic name given to the entity as per schema.org defined within the core context as https://uri.etsi.org/ngsi-ld/name

  • alternateName: An alternative mnemonic name given to the entity as per schema.org

  • description: A textual description of the entity as per schema.org defined within the core context as https://uri.etsi.org/ngsi-ld/description
  • dataProvider: A name identifying the entity providing the data.

See for example the Building Data Model.

Similarly most of the Smart Data Models adopt the Location common definitions, and thus include the following attributes:

  • address: the civic address of the entity as per schema.org.
  • location: the GeoJSON representation of the entity location.

See for example the Building Data Model.

GeoJSON supports the definition of quite complex geometries (Point, LineString, Polygon, MultiPoint, MultiLineString, and MultiPolygon) representing the localisation of entities. Examples of GeoJSON geometries are available here.

General tips

How to specify units of measurement

  1. If your data use the default unit defined in the Data Model, you don't need to specify any. It is implied.
  2. Unless explicitly stated otherwise, all Smart data models use the metric system of measurements by default. Regardless the model specification include explicit reference to the scale adopted.
  3. If your data use a different unit, you will need to use the unitCode metadata annotation in your data (and you will need to adopt the normalised representation). Code used should be the ones defined by UN/CEFACT. E.g.:
"length": {
    "value": 11,
    "metadata": {
        "unitCode": {
            "value": "FOT"
        }
    }
}

How to attach specific time of update to a single attribute

Often, different attributes of the same entity are updated at different points in time as the result of data aggregation across measurements from different sensors. Note that the dateModified value only reflects only the point in time the NGSI v2 implementation received an entity update, it does not correspond to the moment in time at which an attribute changed. In order to be able to define when a specific attribute changed you should you should use the timestamp metadata annotation. E.g.:

"speed": {
    "value": 11,
    "metadata": {
        "timestamp": {
            "value": "2016-03-15T11:00:00.000"
        }
    }
}

Proper usage of relations among Data Models

With the introduction of NGSI-LD, a growing number attributes are being defined which represent relationships between entities.

Previously the Smart data model defined the name of such attributes was normed to be ref followed by the name of the Entity Type referenced by the attribute, for example in the case of Entity Type Device, the full attribute would be refDevice.

With the introduction of NGSI-LD, following the link with practises in the ontologies, such attributes are defined with the usage of a verb (plus optionally an object) such as hasStop, operatedBy, hasTrip, etc.

While in the first case, the entity type to be looked for is implicitly defined by the attribute name, this is no longer the case with NGSI-LD, that's why it is important to use NGSI-LD URNs that convey the type of the target entity, for instance urn:ngsi-ld:gtfs:Stop:S123.

Currently it is up to the logic of your application to include the logic to navigate entity relationships, as it is not yet defined within implementations of the NGSI v2 or NGSI-LD specification.

TODO

  • Add paragraph Validate your own examples