Auto-generate Python APIs from JSON schema specifications

JSON Schema is a vocabulary that allows you to annotate and validate JSON documents.

schemapi is a package that lets you auto-generate simple Python object-based APIs given a valid JSON schema specification.

The motivating case for this package was the Altair visualization library: Altair is a Python API built on the Vega-Lite grammar of visualization, and the bulk of the Altair package is generated automatically using schemapi.

As a very simple example, imagine you have the following simple JSON schema, defined as a Python dictionary:

>>> schema = {
...   'properties': {
...     'name': {'type': 'string'},
...     'age': {'type': 'integer'}
...   }
... }

This schema specifies that a data instance is valid if it has a key "name" that maps to a string, and a key "age" that maps to an integer. So, for example, this dictionary would be valid:

>>> valid = {'name': 'suzie', 'age': 32}

while this dictionary would not:

>>> invalid = {'name': 'suzie', 'age': 'old'}

In Python, you can use the jsonschema package to validate any data objects against this schema. For example, this data passes:

>>> jsonschema.validate(valid, schema)

While this data fails, as indicated by the ValidationError:

>>> jsonschema.validate(invalid, schema)
ValidationError: 'old' is not of type 'integer'

The schemapi package lets you generate a Python API that allows you to build up this kind of data not with raw dictionaries, but with an object-oriented Python approach.

For example, here is how you can create a local module named myschema that includes an object hierarchy designed for creating and validating data under this schema:

>>> import schemapi

>>> api = schemapi.SchemaModuleGenerator(schema, root_name='Person')
>>> api.write_module('myschema.py')
'/Users/jakevdp/myschema.py'

The result of this is that a new Python module named myschema.py is written to disk in the local directory; we can import the root object and use it to construct some data:

>>> from myschema import Person

>>> person = Person(name='suzie', age=32)

This data can be output in the form of a JSON dict:

>>> person.to_dict()
{'age': 32, 'name': 'suzie'}

The object can also be instantiated from a dict:

person = Person.from_dict({'age': 32, 'name': 'suzie'})

The object allows data to be modified in-place using attribute access:

>>> person.name = 'frank'

>>> person.to_dict()
{'age': 32, 'name': 'frank'}

When the object is created, its entries are validated using JSONSchema to ensure that they have the correct type:

>>> Person(name='Bob', age='old')
SchemaValidationError: Invalid specification

        myschema.Person->age, validating 'type'

        'old' is not of type 'integer'

By utilizing JSONSchema definitions and references, much more complicated nested object hierarchies are possible, and the generated classes can be subclassed in order to create domain-specific APIs for specifying data that can be serialized to and from JSON. For an example of a much more complicated schema in action, see the Altair project.

If you do not wish to write a module to disk before importing it, you can construct the module dynamically:

>>> import schemapi
>>> api = schemapi.SchemaModuleGenerator(schema, root_name='Person')
>>> dynamic_module = api.import_as('dynamic_module')

The module returned by this method can be used directly, or you can import from it as with any Python module.

>>> from dynamic_module import Person
>>> person = Person(name='suzie', age=32)
>>> person.to_dict()
{'age': 32, 'name': 'suzie'}

Note, however, that the module lives only in memory, so it will only be available in the Python session in which it is defined.

You can install the released version from PyPI using pip:

$ pip install schemapi

To install the bleeding-edge version from source, you can download this repository and installing locally:

$ git clone https://github.com/altair-viz/schemapi.git
$ cd schemapi
$ pip install .

To run the test suite you must have py.test installed. To run the tests, use

py.test --pyargs schemapi

(you can omit the --pyargs flag if you are running the tests from a source checkout).

schemapi is released under a 3-Clause BSD License.

We welcome any input, feedback, bug reports, and contributions via schemapi's GitHub repository.