Installation

pip install json-converter

Usage

from json_converter.json_mapper import JsonMapper

JSON Mapping

The JsonMapper class is a tool for translating/converting a JSON document into another JSON document with a different structure. The mapping process follows a dictionary-based specification of how fields map to the new JSON format. The main function in the JsonMapper is map that takes a structured specification:

    JsonMapper(json_document).map(specification)

Mapping Specification

The general idea is that the specification describes the resulting structure of the converted JSON document. The dictionary-based specification will closely resemble the schema of the resulting JSON.

Field Specification

A field specification is defined by a list of parameters, the first of which is a name that refers to a field in the current JSON to be converted. This is the only required field.

    <converted_field>: [<original_field>]

For example, given the sample JSON document,

    {
        "person_name": "Juan dela Cruz"
        "person_age": 37 
    }

the simplest mapping that can be done is to translate to a different field name. For example, to map person_name to name in the resulting JSON, the following specification is used:

    {
        'name': ['person_name']
    }

Field Chaining

JSON mapping also supports chaining of fields on either or both side of the specification. For example, using the following specification to the JSON above,

    {
        'person.name': ['person_name'],
        'person.age': ['person_age']
    }

will result in the conversion:

    {
        "person": {
            "name": "Juan dela Cruz",
            "age": 37
        }
    }

To convert back to the original JSON in the example, just reverse the field specification, for example, 'person_name': ['person.name']. Field chaining can be done on multiple levels. However, at the time of writing, JsonMapper does not support direct field chaining for JSON array types. Processing such fields can be expressed through anchoring and nesting.

Post-Processing Using Generic Functions

The JSON mapper allows post processing of field values for more complex translation rules. This is done by specifying a generic Python function that takes an arbitrary list of arguments (*args). The post-processing function is specified after the original field name in the field specification:

    <converted_field>: [<original_field>, <post_processor>{, <args>}*]

JsonMapper will pass the value of the specified field as the first argument to the post-processor. Taking the same example in the previous section, a boolean field adult can be added using this feature. The following spec demonstrates how this can be done:

    {
        'name': ['person_name'],
        'age': ['person_age'],
        'adult': ['person_age', is_adult]
    }

with the post-processor defined as:

    def is_adult(*args):
        age = args[0]
        return age >= 18

At the time of writing, there are a few built-in post-processors that can be applied to some common use-cases like defaulting values (default_to), concatenating lists (concatenate_list), etc. These can be found in post_process.py module.

Anchoring

While the JsonMapper has support for field chaining, for complex JSON with several levels of nesting, combined with long field names and field list, repetitively providing full field chain can be tedious. To be able to express this more concisely, anchoring can be used. Anchoring specifies the root of the JSON structure to map to a new JSON format, relative to the actual root of the original JSON document.

The on Parameter

The map function in the JsonMapper takes a parameter named on that can be used to specify the root of the JSON on which to start mapping. For example, given the following JSON,

    {
        "user": {
            "settings": {
                "basic": {...},
                "advanced": {
                    "security": {
                        "javascript_enabled": true,
                        "allow_trackers": false
                    }
                }
            }
        }
    }

the processing can be anchored on user.settings.advanced.security to translate the security settings. The following specification,

    {
        'javascript': ['javascript_enabled'],
        'trackers': ['allow_trackers']
    }

applied to the JSON above using map(specification, on='user.settings.advanced.security'), will result in,

    {
        "javascript": true,
        "trackers": false
    }

Without the anchor, it's necessary to always include user.settings.advanced.security in the field specification, so the javascript mapping would look like 'javascript': ['user.settings.advanced.security.javascript_enabled'].

The $on Specification

Another way of specifying the anchoring field is by directly adding it to the specification using the $on keyword. Unlike field specifications, the $on keyword takes a plain string and not a list/vector. For example, the previous sample specification can be alternatively expressed as,

    {
        '$on': 'user.settings.advanced.security',
        "javascript": ['javascript_enabled'],
        "trackers": ['allow_trackers']
    } 

Using this specification, the mapping can be invoked without the on parameter. Keywords in specifications are, at the time of writing, case-sensitive, so $On, $ON, etc. are not recognised.

Chaining on and $on

The on parameter and the $on keyword do not override, but instead are chained together. The existence of both during a mapping call results in the $on field chain being concatenated to the value provided in through the on parameter. For example, the following invocation is equivalent to the previous two above:

    map({
        '$on': 'advanced.security',
        "javascript": true,
        "trackers": false
    }, on='user.settings')

The user.settings field supplied through on parameter will be treated as a prefix to the advanced.security field specified through the $on keyword.

Nested Specification

Aside from field specifications, nested dictionary-like specification can be provided to any recognised fields in the root specification. Nesting is useful for expressing nesting on single objects, or for applying conversion to a list of JSON objects defined in an array.

Single Object Nesting

For single objects, nested specs can be defined to look like the resulting JSON object. Nesting specification this way is a more expressive alternative to field chaining that was demonstrated above. For example, the following JSON, similar to the previous sections,

    {
        "person_name": "Jane Eyre",
        "person_age": 30
    }

can be mapped using nested specification defined with a nested person object:

    {
        'person': {
            'name': ['person_name'],
            'age': ['person_age']
        }
    }

Anchoring in nested specifications is also supported. However, unlike anchors in the main specification that can be expressed through the on parameter, nested anchors can only be specified using the $on keyword. It is also important to note that nested anchors are defined relative to the parent specification. For example, the following JSON,

    {
        "product_info" {
            "manufacturing": {
                "location": "Cambridge, UK", 
                "manufacturing_date": "2020-03-05",
                "best_by_date": "2020-09-05"
            }
        }
    }

can be mapped using the following nested specification,

    {
        '$on': 'product_info',
        'production': {
            '$on': 'manufacturing',
            'date': ['manufacturing_date']
        }
    }

This mapping will result in the following JSON:

    {
        "production": {
            "date": "2020-03-05"
        }
    }

Applying Specification to JSON Arrays

The JSON mapping utility can distinguish between JSON object nodes and JSON array, and applies specification accordingly. When it determines that a field referred to by the specification is a collection of JSON objects, it applies the rules to each one of them iteratively. Note, however, that, as earlier mentioned in this document, using field chaining to refer to nested JSON array is not supported at this time. To apply specifications to JSON arrays, they need to be explicitly anchored if they are nested within the original JSON document.

To illustrate, the following JSON object,

    {
        "books": [
            {
                "title": "A Python Book",
                "price": 23.75
            },
            {
                "title": "A Novel",
                "price": 7.99
            },
            {
                "title": "Compilation of Fun Stuff",
                "price": 10.10
            }
        ]
    }

can be translated using the following specification (assume translate and convert are defined; see post-processing for more information on this),

    {
        '$on': 'books',
        'titulo': ['title', translate, 'es'],
        'precio': ['price', convert, 'eur']
    }

Notice that, since the specification is anchored to the books node, only the list of field specifications are defined. Specifications applied to multiple objects this way are expressed as if it applies to a single object. This sample translation will return an array of JSON objects and not a JSON object containing an array. If a nested array is desired, the specification above can be nested instead:

    {
        'libros': {
            '$on': 'books',
            'titulo': ['title', translate, 'es'],
            'precio': ['price', convert, 'eur']
        }
    }

Filtering

When working on collections of data, it is sometimes required to only process some based on some criteria. For JsonMapper, this is done by using filtering in the specification. Filters take the form of a field specification with the second argument being a boolean function, also called a predicate.

    '$filter': [<original_field>, <predicate_function>{, args}*]

Just like in generic functions for post-processing, JsonMapper will pass the value of found in the original_field as the first parameter along with the rest of args if they exist.

For example, to process only books whose prices are above 10.00 from the sample books JSON above, the following spec can be used:

    {
        'expensive_books': {
            '$on': 'books',
            '$filter': ['price', greater_than, 10],
            'book_title': ['title'],
            'book_price': ['price']
        }
    }

with the predicate greater_than defined as,

    def greater_than(*args):
        number = args[0]
        other_number = args[1]
        return number > other_number

While filtering can be applied to single JSON nodes, the application can be limited. Any JSON object filtered out, will appear as an empty JSON object in the resulting document.

JSON Literals

There are situations when the resulting JSON need to contain fields and values outside the scope of the source JSON document. In such cases, it's possible to define a post-processor that plugs-in a pre-defined dictionary-like or list structure. However, JsonMapper also provides support for including literals in the specification.

Using Keywords

As mentioned above, there are 2 types of node that can be used for adding predefined values into the specification, which are object, and array. To specify a JSON object literal as field value in the resulting JSON document, the $object keyword is used with a dictionary-like structure:

    <field_name>: ['$object', <object_value>]

For example:

    'metadata': ['$object', {
        'date_created': '2020-03-13',
        'author': 'Jane Doe'
    }]

For collections or list of JSON objects, the '$array' is used instead:

    <field_name>: ['$array': <list>]

For example:

    'authors': ['$array', [
        {
            'name': 'Peter Z',
            'institution': 'Some University'
        },
        {
            'name': 'Mary Q',
            'institution': 'Some Research Institute'
        }
    ]]

Convenience Methods

from json_converter.json_mapper import json_object, json_array

For convenience, the json_mapper module makes available 2 helper methods that allow easy inclusion of JSON predefined JSON nodes into the specification. These are json_object for JSON objects, and json_array for JSON arrays. json_object expects a dictionary-like structure as an input.

For example, the previous example can be expressed as the following:

    'metadata': json_object({
        'date_created': '2020-03-13',
        'author': 'Jane Doe'
    })

The json_array method treats argument list as a list of JSON objects. For example:

    'authors': json_array(
        {
            'name': 'Peter Z',
            'institution': 'Some University'
        },
        {
            'name': 'Mary Q',
            'institution': 'Some Research Institute'
        }
    )

Note that any list literal provided within the json_array method is treated as a single object. For instance, the call json_array([{'object_id': 123}, {'object_id': 456}]) has one item in the resulting list of list.