OCSF Schema Tools

Tools for working with the OCSF schema in Python.

About

This project began with two goals:

1. Provide the OCSF community with a validator that tests for breaking changes in ocsf-schema PRs.
2. Begin to provide the OCSF community with more composable tools and libraries, as well as approachable reference implementations of OCSF related functions, in order to make OCSF more "hackable."

The scope of this project may grow to include things like a reference implementation OCSF schema compiler.

The project targets Python 3.11 for a balance of capability and availability, is divided into discrete packages.

The Schema Package

The ocsf_tools.schema package contains Python data classes that represent an OCSF schema as compiled and exported by the OCSF server's export/schema endpoint.

It also includes utilities to parse the schema from a JSON string or file, as well as a lightweight HTTP client that can retrieve a version of the schema over HTTP and cache it on the local filesystem.

The easiest way to instantiate a schema is with the get_schema function, which accepts a filename or a semver.

schema = get_schema("1.1.0")
schema = get_schema("./1.3.0-dev.json")

The Compare Package

The ocsf_tools.compare package compares two versions of the OCSF schema and generates a type safe difference. Its aim is to make schema comparisons easy to work with.

This package grew out of a library used internally at Query. The original is used extensively to manage upgrading Query's data model to newer versions of OCSF, as well as to build adapters between different OCSF flavors (like AWS Security Lake on rc2 and Query on 1.1).

There is a very simple __main__ implementation to demonstrate the comparison. You can use it as follows:

$ poetry run python -m ocsf_tools.compare 1.0.0 1.2.0

The comparison API is straightforward. Want to look for removed events?

diff = compare(get_schema("1.0.0", "1.1.0"))
for name, event in diff.classes.items():
    if isinstance(event, Removal):
        print(f"Oh no, we've lost {name}!")

Or changed data types?

diff = compare(get_schema("1.0.0", "1.1.0"))
for name, event in diff.classes.items():
    if isinstance(event, ChangedEvent):
        for attr_name, attr in event.attributes.items():
            if isinstance(attr, ChangedAttr):
                if isinstance(attr.type, Change):
                    print(f"Who changed this data type? {name}.{attr_name}")

Or new objects?

diff = compare(get_schema("1.0.0", "1.1.0"))
for name, obj in diff.objects.items():
    if isinstance(obj, Addition):
        print(f"A new object {name} has been discovered!")

The Validation Package

The ocsf_tools.validation package provides a lightweight framework for validators. It was inspired by the needs of ocsf-validator, which may be ported to this framework in the future.

The Compatibility Package

The ocsf_tools.compatibility package uses the other packages to provide a validator that compares two OCSF schema and looks for breaking changes as defined in the OCSF FAQ.

The compatibility validator can be run as follows:

$ poetry run python -m ocsf_tools.compatibility --before 1.0.0 --after 1.2.0 --cache ./cache

Getting Started

PyPI

The easiest way to install ocsf-tools is from PyPI using pip or poetry:

$ pip install ocsf-tools

NOTE: This is not actually published to PyPI yet.

From Source

If you want to work with the source, the recommended installation is with asdf and poetry.

$ asdf install
$ poetry install

Contributing

This project uses ruff for formatting and linting, pyright for type checking, and pytest as its test runner.

Before submitting a PR, make sure you've run following:

$ poetry run ruff format
$ poetry run ruff check
$ poetry run pyright
$ poetry run pytest

Type Checking

With great effort, this library passes pyright's strict mode type checking. Keep it that way! The OCSF schema is big, and even the metaschema is a lot to hold in your head. Having the type checker identify mistakes for you can be very helpful.

There is one cast used from the concrete ChangedModel types (ChangedSchema, ChangedAttr, etc.) in the compare package to the generic type. For the life of me, I can't figure it out. I blame pyright but it's probably my own fault.

Tests

Running unit tests:

$ poetry run pytest -m "not integration"

Running integration tests:

$ poetry run pytest -m integration

NOTE: Some of the integration tests require an OCSF server instance, and are using the public instance at https://schema.ocsf.io. This should probably use a local instance of the OCSF server instead.