Skip to main content

A package to convert Python type annotations into JSON schemas

Project description



Package that uses static analysis - ast - to convert Python 3 function type annotations to JSON schemas.

This allows you to auto-generate the validation schemas for JSON-RPC backend functions written in Python.

Current support is for Python 3.8+ and JSON schema draft 7+.

Getting started


From a Python 3.8+ environment, run pip install pytojsonschema.

Scan a package

After installing the package, you can open a python terminal from the root of the repo and run:

import os
import pprint

from pytojsonschema.functions import process_package

pprint.pprint(process_package(os.path.join("test", "example")))

The example package will be scanned and JSON schemas will be generated for all the top level functions it can find.

Scan a file

You can also target specific files, which won't include the package namespacing in the result value. Following on the same terminal:

from pytojsonschema.functions import process_file

pprint.pprint(process_file(os.path.join("test", "example", "")))

Include and exclude patterns

Include and exclude unix-like patterns can be used to filter function and module names we want to allow/disallow for scanning. See the difference when you now run this instead:

pprint.pprint(process_package(os.path.join("test", "example"), exclude_patterns=["_*"]))

Similarly, but applied to specific files:

pprint.pprint(process_file(os.path.join("test", "example", ""), exclude_patterns=["_*"]))

Things to take into account:

  • Exclude pattern matching overwrite include matches.
  • files are not affected by pattern rules and are always scanned. However, you can still filter its internal functions.

Type annotation rules

Fitting Python's typing model to JSON means not everything is allowed in your function signatures. This is a natural restriction that comes with JSON data serialization. Hopefully, most of the useful stuff you need is allowed.

Allowed types

Base types

Basic types bool, int, float, str, None and typing.Any are allowed. Also, you can build more complex, nested structures with the usage of typing.Union, typing.Optional, typing.Dict (Only str keys are allowed) and typing.List. All these types have a direct, non-ambiguous representation in both JSON and JSON schema.

Custom types

Your functions can also use custom types like the ones defined using an assignment of typing.Union, typing.List, typing.Dict and typing.Optional, as in:

ServicePort = typing.Union[int, float]
ServiceConfig = typing.Dict[str, typing.Any]

You can use one of the new Python 3.8 features, typing.TypedDict, to build stronger validation on dict-like objects (Only class-based syntax). As you can see, you can chain types with no restrictions:

class Service(typing.TypedDict):
    address: str
    port: ServicePort
    config: ServiceConfig
    tags: typing.List[str]
    debug: bool = False

Note: Whilst Python itself will not auto-populate default values, you can use them to make the property not required.

Also, if you need to restrict the choices for a type, you can use Python enums:

import enum

class HTTPMethod(enum.Enum):
    GET = "GET"
    POST = "POST"

def my_func(http_method: HTTPMethod):
    pass  # My code

Note: All properties of the enum must be constants: None, int, float, bool, str.

Note: The resulting validation uses the different enum values, e.g. HTTPMethod.GET.value or simply "GET", as the valid choices and not the enum instance itself, e.g. HTTPMethod.GET, as that is what JSON schema can understand. This can lead to some incompatibilities with other static analysis tools like, so bear that in mind.

Importing types from other files

You can import these custom types within your package and they will be picked up. However, due to the static nature of the scan, custom types coming from external packages can't be followed and hence not supported. In other words, you can only share these types within your package, using relative imports.

Other static analysis tools like mypy use a repository with stub files to solve this issue, see This is out of the scope for a tiny project like this, at least for now.


  1. The functions you want to scan need to be type annotated. Kind of obvious requirement, right?

  2. Only the types defined in the previous section can be used. They are the types that can be safely serialised as JSON.

  3. Function arguments are meant to be passed in key-value format, like a json object. This puts a couple of restrictions regarding *args, **kwargs, positional-only and keyword-only arguments:

    The following is allowed:

    • **kwargs: def func(**kwargs): pass
    • keyword-only arguments: def func(*, a): pass

    The following is not allowed:

    • *args: def func(*args): pass
    • positional-only arguments: def func(a, /): pass

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for pytojsonschema, version 1.11.1
Filename, size File type Python version Upload date Hashes
Filename, size pytojsonschema-1.11.1-py3-none-any.whl (11.7 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size pytojsonschema-1.11.1.tar.gz (15.5 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page