.. _json_converter:

JSON Converter
==============

The JSON converter handles JSON files. To use the JSON converter, specify ``type = "json"`` in your converter
definition.

Configuration
-------------

The JSON converter supports parsing multiple JSON documents out of a single file.
In order to support JSON path expressions, each JSON document is fully parsed into memory.
For large documents, this may take considerable time and memory. Thus, it is usually better to have multiple
smaller JSON documents per file, when possible.

Since a single JSON document may contain multiple features, the JSON parser supports a
`JSONPath <http://goessner.net/articles/JsonPath/>`__ expression pointing to each feature element. This can
be specified using the ``feature-path`` element.

The ``fields`` element in a JSON converter supports two additional attributes, ``path`` and ``json-type``.
``path`` should be a `JSONPath <http://goessner.net/articles/JsonPath/>`__ expression, which is relative to the
``feature-path``, if defined (above). For absolute paths, ``root-path`` may be used instead of ``path``.
``json-type`` should specify the type of JSON field being read. Valid values are: **string**, **float**, **double**,
**integer**, **long**, **boolean**, **geometry**, **array** and **object**. The value will be appropriately typed,
and available in the ``transform`` element as ``$0``. Geometry types can handle either WKT strings or GeoJSON
geometry objects.

.. _json_converter_functions:

JSON Transform Functions
------------------------

The ``transform`` element supports referencing the JSON element through ``$0``. Each column will initially
be typed according to the field's ``json-type``. Most types will be converted to the equivalent Java class,
e.g. java.lang.Integer, etc. **array** and **object** types will be raw JSON elements, and thus usually
require further processing (e.g. ``jsonList`` or ``jsonMap``, below).

In addition to the standard functions in :ref:`converter_functions`, the JSON converter provides the following
JSON-specific functions:

jsonToString
~~~~~~~~~~~~

This will convert a JSON element to a string. It can be useful for quickly representing a complex object, for
example in order to create a feature ID based on the hash of a row.

jsonList
~~~~~~~~

This function converts a JSON array element into a java.util.List. It requires two parameters; the first is the
type of the list elements as a string, and the second is a JSON array. The type of list elements must be one
of the types defined in :ref:`attribute_types`. See below for an example.

jsonMap
~~~~~~~

This function converts a JSON object element into a java.util.Map. It requires three parameters; the first is the
type of the map key elements as a string, the second is the type of the map value elements as a string, and the
third is a JSON object. The type of keys and values must be one of the types defined in :ref:`attribute_types`.
See below for an example.

mapToJson
~~~~~~~~~

This function converts a java.util.Map into a JSON string. It requires a single parameter, which must be a
java.util.Map. It can be useful for storing complex JSON as a single attribute, which can then be queried
using GeoMesa's JSON attribute support. See :ref:`json_attributes` for more information.

Example Usage
-------------

Assume the following SimpleFeatureType:

::

  geomesa.sfts.example = {
    attributes = [
      { name = "name",    type = "String"          }
      { name = "age",     type = "Integer"         }
      { name = "weight",  type = "Double"          }
      { name = "hobbies", type = "List[String]"    }
      { name = "skills",  type = "Map[String,Int]" }
      { name = "source",  type = "String"          }
      { name = "geom",    type = "Point"           }
    ]
  }

And the following JSON document:

::

  {
    "DataSource": { "name": "myjson" },
    "Features": [
      {
        "id": 1,
        "name": "phil",
        "physicals": {
          "age": 32,
          "weight": 150.2
        },
        "hobbies": [ "baseball", "soccer" ],
        "languages": {
          "java": 100,
          "scala": 70
        },
        "geometry": { "type": "Point", "coordinates": [55, 56] }
      },
      {
        "id": 2,
        "name": "fred",
        "physicals": {
          "age": 33,
          "weight": 150.1
        },
        "hobbies": [ "archery", "tennis" ],
        "languages": {
          "c++": 10,
          "fortran": 50
        },
        "geometry": { "type": "Point", "coordinates": [45, 46] }
      }
    ]
  }

You could ingest with the following converter:

::

 geomesa.converters.myjson = {
   type         = "json"
   id-field     = "$id"
   feature-path = "$.Features[*]"
   fields = [
     { name = "id",      json-type = "integer",  path = "$.id",               transform = "toString($0)"                }
     { name = "name",    json-type = "string",   path = "$.name",             transform = "trim($0)"                    }
     { name = "age",     json-type = "integer",  path = "$.physicals.age",                                              }
     { name = "weight",  json-type = "double",   path = "$.physicals.weight"                                            }
     { name = "hobbies", json-type = "array",    path = "$.hobbies",          transform = "jsonList('string', $0)"      }
     { name = "skills",  json-type = "map",      path = "$.languages",        transform = "jsonMap('string','int', $0)" }
     { name = "geom",    json-type = "geometry", path = "$.geometry",         transform = "point($0)"                   }
     { name = "source",  json-type = "string",   root-path = "$.DataSource.name"                                        }
   ]
 }