JSON-NTV is a universal representation format. It allows the sharing and conversion of any type of data.

The NTV format is part of the Environmental Sensing Project

For more information, see the user guide or the github repository.

The semantic level of shared JSON (or CSV) data (e.g. Open Data) remains low, which makes automated reuse difficult.

JSON-NTV proposes to enrich it to obtain a real interoperable exchange format.

The NTV format consists of representing data by three attributes: a name, a type and a value. This representation is common in programming languages (for example a variable with Python typing is defined by age: int = 25), however the JSON format represents data with only a value or a key:value pair.

The JSON-NTV extension consists of including the type in the name to associate it with a value (for example {'age:int': 25} is the JSON representation of the NTV triplet ('age', 'int' , 25 ) ).

This approach makes it possible to reversibly represent any simple or complex data by a JSON structure (high interoperability).

In [1]: from shapely.geometry import Point
        from datetime import date
        from pprint import pprint

In [2]: pprint(Ntv.obj(21).expand())
Out[2]: {'name': '', 'type': 'json', 'value': 21}

In [3]: pprint(Ntv.obj({"paris:point": [2.3, 48.9] }).expand())
Out[3]: {'name': 'paris', 'type': 'point', 'value': [2.3, 48.9]}

In [4]: pprint(Ntv.obj({"cities::point": [[2.3, 48.9], [4.8, 45.8] }).expand())
Out[4]: {'name': 'cities',
         'type': 'point',
         'value': [{'name': '', 'type': 'point', 'value': [2.3, 48.9]},
                   {'name': '', 'type': 'point', 'value': [4.8, 45.8]}]}

In [5]: pprint(Ntv.obj({"paris:point": [2.3, 48.9], "start:date": "2023-08-03", "measurement": 45.8}).expand())
Out[5]: {'name': '',
         'type': '',
         'value': [{'name': 'paris', 'type': 'point', 'value': [2.3, 48.9]},
                   {'name': 'start', 'type': 'date', 'value': '2023-08-03'},
                   {'name': 'measurement', 'type': 'json', 'value': 45.8}]}

Note: This typing syntax can also be used for CSV file headers

With this approach, two NTV entities are defined:

• a primitive entity which is not composed of any other entity (NTV-single),
• a structured entity which is an ordered sequence of NTV entities (NTV-list).

as well as two JSON formats:

• simple format when the name and the type are not present (e.g. 25),
• named format when the name or type is present ((e.g. {'age': 25} or {':int': 25})).

The type incorporates a notion of namespace that can be nested.

For example, the type: ns1.ns2.type_a means that:

• ns1. is a namespace defined in the global namespace,
• ns2. is a namespace defined in the ns1. namespace.,
• type_a is defined in the ns2. namespace.

The type can be extended with additional data.

For example, the type: float[kg] means that:

• float is a main type of the data,
• kg is an additional type (e.g. an unit),
• float[kg] may represent a Quantity.

This structuring of type makes it possible to reference any type of data that has a JSON representation and to consolidate all the shared data structures within the same tree of types.

Several variations and use cases of the NTV format are defined:

• Tabular and multidimensional data exchange format (e.g. open-data)
• Compact, reversible and semantic pandas-JSON interface
• Comment and change management of JSON data
• visualization of JSON or NTV tree
• JSON data editor
• data validation (value conformity to the Datatype)

The flowchart below explain how to convert and exchange native entities through NTV and JSON format.

flowchart LR
    text["#10240;#10240;JSON#10240;#10240;\ntext"]
    val["#10240;JSON-NTV#10240;\nvalue"]
    ntv["#10240;#10240;#10240;NTV#10240;#10240;#10240;\nentity"]
    nat["#10240;native#10240;\nentity"]
    text--->|JSON load|val
    val--->|JSON dump|text
    val--->|NTV from JSON|ntv
    ntv--->|from NTV|nat
    ntv--->|NTV to JSON|val
    nat--->|to NTV|ntv

The conversion between native entity and JSON-text is reversible (lossless round trip).

In [6]: loc_and_date = {'newyear': date(2023, 1, 2), 'Paris': Point(2.3, 48.9)}
        json_loc_date = Ntv.obj(loc_and_date).to_obj(encoded=True)
        print(json_loc_date, type(json_loc_date))
Out[6]: {"newyear:date": "2023-01-02", "Paris:point": [2.3, 48.9]} <class 'str'>

In [7]: Ntv.obj(json_loc_date).to_obj(format='obj') == loc_and_date
Out[7]: True

Properties :

• each NTV object has a unique JSON representation
• each JSON data corresponds to a unique NTV entity
• an NTV entity is a tree where each node is an NTV entity and each leaf an NTV-Single entity
• an NTV entity is a neutral representation (independent of a software or hardware platform)