JSON Schema

The JSONSchema class represents a compiled JSON schema document. Once instantiated, a JSONSchema object contains all the structural information and executable code it requires for evaluating JSON instances, and it may be re-used any number of times for such evaluation. References to other schemas are resolved during construction, guaranteeing further that any referenced schemas are also fully loaded, compiled and ready for use and re-use.

JSONSchema is a specialization of the JSON class, and provides all the capabilities of its ancestor, as described in the JSON guide. Only its JSON type is limited – to one of "object" and "boolean" – in accordance with the JSON Schema specification. As we might expect, JSONSchema introduces several new properties and behaviours, which we’ll explore in the following sections.

Initializing the catalog

Before we can begin creating and working with schemas, we must set up a catalog. For the examples shown on the remainder of this page, we’ll use the following:

>>> from jschon import create_catalog
>>> catalog = create_catalog('2020-12')

Creating a schema

There are several different ways to instantiate a JSONSchema:

  • Create it directly from a schema-compatible Python object such as a dict or a bool.

  • Deserialize it from a JSON file or a JSON string using the loadf() or loads() class method.

  • Retrieve it from the Catalog by providing a schema URI, which maps to a schema file on disk.

But first, let’s import the classes that we’ll be using:

>>> from jschon import JSONSchema, URI

Creating a schema from a Python object

A JSONSchema object can be created directly from any schema-compatible Python object, such as a dict or a bool.

For boolean schemas and empty schemas, it’s as simple as:

>>> true_schema = JSONSchema(True)
>>> false_schema = JSONSchema(False)
>>> empty_schema = JSONSchema({})

Creating a JSONSchema object from any non-empty Mapping, however, requires that we specify a meta-schema. Here’s a very basic example of a schema that simply ensures that a JSON value represents an integer:

>>> int_schema = JSONSchema({
...     "type": "integer"
... }, metaschema_uri=URI("https://json-schema.org/draft/2020-12/schema"))

The metaschema_uri parameter resolves to a Metaschema object, which provides the new JSONSchema instance with Keyword classes for each of its keywords. The meta-schema URI may be parameterized, as above, or it may be provided using the "$schema" keyword:

>>> int_schema = JSONSchema({
...     "type": "integer",
...     "$schema": "https://json-schema.org/draft/2020-12/schema"
... })

If both are provided, the "$schema" keyword takes precedence:

>>> int_schema = JSONSchema({
...     "type": "integer",
...     "$schema": "https://json-schema.org/draft/2020-12/schema"
... }, metaschema_uri=URI("https://json-schema.org/draft/2019-09/schema"))
>>> int_schema.metaschema_uri

An identifying URI is automatically generated for every root schema:

>>> int_schema.uri

The schema URI is used as the key for caching the schema in the catalog, and is required for resolving references to itself and to any subschemas it may contain. If the schema is intended to be referenced from other schemas in the catalog, then a URI should be provided explicitly. This may either be passed via the uri parameter to the constructor, or declared in the schema document itself using the "$id" keyword. If both are provided, the "$id" keyword takes precedence:

>>> int_schema = JSONSchema({
...     "type": "integer",
...     "$schema": "https://json-schema.org/draft/2020-12/schema",
...     "$id": "https://example.com/the-real-id"
... }, uri="https://example.com/not-the-real-id")
>>> int_schema.uri

Deserializing a schema from a JSON text document

Suppose that we have a string containing a JSON schema, such as the following, which validates a numeric JSON value:

>>> schema_text = '''{
...     "$schema": "https://json-schema.org/draft/2020-12/schema",
...     "$id": "https://example.com/num-schema",
...     "type": "number"
... }'''

We can deserialize this JSON text into a new JSONSchema instance using the loads() class method:

>>> num_schema = JSONSchema.loads(schema_text)

If the schema is stored in a JSON text file, we can deserialize directly from the file by providing the file path to the loadf() class method:

>>> num_schema = JSONSchema.loadf('/path/to/num-schema.json')

The argument to loadf() may be a plain str, or any PathLike object; for example:

>>> import pathlib
>>> schema_path = pathlib.Path(__file__).parent / 'num-schema.json'
>>> num_schema = JSONSchema.loadf(schema_path)

Both loads() and loadf() accept keyword arguments that are passed through to the JSONSchema constructor, in case we need to parameterize the meta-schema URI, the schema URI, or any other JSONSchema constructor argument:

>>> num_schema = JSONSchema.loads(schema_text, metaschema_uri=URI("https://json-schema.org/draft/2020-12/schema"))
>>> num_schema = JSONSchema.loadf(schema_path, uri=URI("https://example.com/num-schema"))

Retrieving a schema from the catalog

Finally, a JSONSchema object may be instantiated implicitly, when retrieving it by URI from the Catalog. If the schema is not already cached, it is loaded from disk and compiled on the fly. This approach requires the catalog to be configured with an appropriate base URI-to-directory mapping. For more information, see Reference loading.