Developer Guide

This document describes the internal architecture and main concepts behind validate-pyproject and targets contributors and plugin writers.

How it works

validate-pyproject relies mostly on a set of specification documents represented as JSON Schema. To run the checks encoded under these schema files validate-pyproject uses the fastjsonschema package.

This procedure is defined in the api module, specifically under the Validator class. Validator objects use SchemaRegistry instances to store references to the JSON schema documents being used for the validation. The formats module is also important to this process, since it defines how to validate the custom values for the "format" field defined in JSON Schema, for "string" values.

Checks for PEP 517, PEP 518 and PEP 621 are performed by default, however these standards do not specify how the tool table and its subtables are populated.

Since different tools allow different configurations, it would be impractical to try to create schemas for all of them inside the same project. Instead, validate-pyproject allows Plugins to provide extra JSON Schemas, against which tool subtables can be checked.

Plugins

Plugins are a way of extending the built-in functionality of validate-pyproject, can be simply described as functions that return a JSON schema parsed as a Python dict:

def plugin(tool_name: str) -> dict:
    ...

These functions receive as argument the name of the tool subtable and should return a JSON schema for the data structure under this table (it should not include the table name itself as a property).

To use a plugin you can pass an extra_plugins argument to the Validator constructor, but you will need to wrap it with PluginWrapper to be able to specify which tool subtable it would be checking:

from validate_pyproject import api


def your_plugin(tool_name: str) -> dict:
    return {
        "$id": "https://your-urn-or-url",  # $id is mandatory
        "type": "object",
        "description": "Your tool configuration description",
        "properties": {
            "your-config-field": {"type": "string", "format": "python-module-name"}
        },
    }


available_plugins = [
    plugins.PluginWrapper("your-tool", your_plugin),
]
validator = api.Validator(extra_plugins=available_plugins)

Please notice that you can also make your plugin “autoloadable” by creating and distributing your own Python package as described in the following section.

If you want to disable the automatic discovery of all “autoloadable” plugins you can pass plugins=[] to the constructor; or, for example in the snippet above, we could have used plugins=... instead of extra_plugins=... to ensure only the explicitly given plugins are loaded.

Distributing Plugins

To distribute plugins, it is necessary to create a Python package with a validate_pyproject.tool_schema entry-point.

For the time being, if using setuptools, this can be achieved by adding the following to your setup.cfg file:

# in setup.cfg
[options.entry_points]
validate_pyproject.tool_schema =
    your-tool = your_package.your_module:your_plugin

When using a PEP 621-compliant backend, the following can be add to your pyproject.toml file:

# in pyproject.toml
[project.entry-points."validate_pyproject.tool_schema"]
your-tool = "your_package.your_module:your_plugin"

The plugin function will be automatically called with the tool_name argument as same name as given to the entrypoint (e.g. your_plugin("your-tool")).

Also notice plugins are activated in a specific order, using Python’s built-in sorted function.