Create loader schemas for validating `kind.yml` files

[ahal]

The format of the kind.yml files is currently determined by loaders. We should provide schemas for what these loaders expect and validate kind.yml files against them. This would result in clearer expectations about what is allowed in these files.

This came from a request from :asuth who is looking into ways of integrating Taskgraph with Searchfox. He'd prefer the schema follow the JSON-schema spec. I'm not sure I'd want two separate schema validation methods being used (jsonschema + voluptuous), but maybe we could have a CI step to export the schema into the JSON-schema format or something.

[ahal]

Fixing this bug will involve creating a new voluptuous schema (like we have for transforms), here: https://github.com/taskcluster/taskgraph/blob/main/src/taskgraph/loader/transform.py

Then somewhere around here in generator.py, we'd evaluate the contents of the kind.yml against said schema. Possibly we could fetch the schema by looking for a common global variable in the loader module, or we could invent some more complex way of registering loaders..

[Aakash017]

Hello @ahal I would like to work on this

[Aakash017]

@ahal As per my understanding we have to create new schema using voluptuous and then evaluate all kind.yml files

[ahal]

Yes, that's correct! Though we'd only evaluate them as they are being loaded (we don't want to evaluate them all in row or anything like that, as they aren't always all needed).

[Aakash017]

@ahal

from taskgraph.transforms.base import TransformSequence
from taskgraph.util.schema import Schema, optionally_keyed_by, resolve_keyed_by

_status_type = Any(
    "on-completed",
    "on-defined",
    "on-exception",
    "on-failed",
    "on-pending",
    "on-resolved",
    "on-running",
)

_recipients = [
    {
        Required("type"): "email",
        Required("address"): optionally_keyed_by("project", "level", str),
        Optional("status-type"): _status_type,
    },
    {
        Required("type"): "matrix-room",
        Required("room-id"): str,
        Optional("status-type"): _status_type,
    },
    {
        Required("type"): "pulse",
        Required("routing-key"): str,
        Optional("status-type"): _status_type,
    },
    {
        Required("type"): "slack-channel",
        Required("channel-id"): str,
        Optional("status-type"): _status_type,
    },
]

_route_keys = {
    "email": "address",
    "matrix-room": "room-id",
    "pulse": "routing-key",
    "slack-channel": "channel-id",
}
"""Map each type to its primary key that will be used in the route."""

NOTIFY_SCHEMA = Schema(
    {
        Exclusive("notify", "config"): {
            Required("recipients"): [Any(*_recipients)],
            Optional("content"): {
                Optional("email"): {
                    Optional("subject"): str,
                    Optional("content"): str,
                    Optional("link"): {
                        Required("text"): str,
                        Required("href"): str,
                    },
                },
                Optional("matrix"): {
                    Optional("body"): str,
                    Optional("formatted-body"): str,
                    Optional("format"): str,
                    Optional("msg-type"): str,
                },
                Optional("slack"): {
                    Optional("text"): str,
                    Optional("blocks"): list,
                    Optional("attachments"): list,
                },
            },
        },
        # Continue supporting the legacy schema for backwards compat.
        Exclusive("notifications", "config"): {
            Required("emails"): optionally_keyed_by("project", "level", [str]),
            Required("subject"): str,
            Optional("message"): str,
            Optional("status-types"): [_status_type],
        },
    },
    extra=ALLOW_EXTRA,
)

print(NOTIFY_SCHEMA({'classic': {'email': 'foo@example.com', 'password': 'bar'}}))

Is it something like this we have to evaluate or something else?