Fix the generated DABs JSON schema (#1322)

## Changes
This PR fixes bundle schema being broken because `for_each_task: null`
was set in the generated schema. This is not valid according to the JSON
schema specification and thus the Red Hat YAML VSCode extension was
failing to parse the YAML configuration.

This PR fixes: https://github.com/databricks/cli/issues/1312

## Tests
The fix itself was tested manually. I asserted that the autocompletion
works now. This was mistakenly overlooked the first time around when the
regression was introduced in https://github.com/databricks/cli/pull/1204
because the YAML extension provides best-effort autocomplete suggestions
even if the JSON schema fails to load.

To prevent future regressions we also add a test to assert that the JSON
schema generated itself is a valid JSON schema object. This is done via
using the `ajv-cli` to validate the schema. This package is also used by
the Red Hat YAML extension and thus provides a high fidelity check for
ensuring the JSON schema is valid.

Before, with the old schema:
```
shreyas.goenka@THW32HFW6T cli-versions % ajv validate -s proj/schema-216.json -d ../bundle-playground-3/databricks.yml
schema proj/schema-216.json is invalid
error: schema is invalid: data/properties/resources/properties/jobs/additionalProperties/properties/tasks/items/properties/for_each_task must be object,boolean, data/properties/resources/properties/jobs/additionalProperties/properties/tasks/items must be array, data/properties/resources/properties/jobs/additionalProperties/properties/tasks/items must match a schema in anyOf
```

After, with the new schema:
```
shreyas.goenka@THW32HFW6T cli-versions % ajv validate -s proj/schema-dev.json -d ../bundle-playground-3/databricks.yml
../bundle-playground-3/databricks.yml valid
```

After, autocomplete suggestions:

<img width="600" alt="Screenshot 2024-03-27 at 6 35 57 PM"
src="https://github.com/databricks/cli/assets/88374338/d0a62402-e323-4f36-854d-332b33cbeab8">
This commit is contained in:
shreyas-goenka 2024-03-28 16:55:36 +05:30 committed by GitHub
parent eea34b2504
commit cddc5f97f8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 29 additions and 1 deletions

View File

@ -89,3 +89,29 @@ jobs:
run: |
# Exit with status code 1 if there are differences (i.e. unformatted files)
git diff --exit-code
validate-bundle-schema:
runs-on: ubuntu-latest
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Setup Go
uses: actions/setup-go@v5
with:
go-version: 1.21.x
# Github repo: https://github.com/ajv-validator/ajv-cli
- name: Install ajv-cli
run: npm install -g ajv-cli@5.0.0
# Assert that the generated bundle schema is a valid JSON schema by using
# ajv-cli to validate it against a sample configuration file.
# By default the ajv-cli runs in strict mode which will fail if the schema
# itself is not valid. Strict mode is more strict than the JSON schema
# specification. See for details: https://ajv.js.org/options.html#strict-mode-options
- name: Validate bundle schema
run: |
go run main.go bundle schema > schema.json
ajv -s schema.json -d ./bundle/tests/basic/databricks.yml

View File

@ -95,7 +95,9 @@ func safeToSchema(golangType reflect.Type, docs *Docs, traceId string, tracker *
// HACK to unblock CLI release (13th Feb 2024). This is temporary until proper
// support for recursive types is added to the schema generator. PR: https://github.com/databricks/cli/pull/1204
if traceId == "for_each_task" {
return nil, nil
return &jsonschema.Schema{
Type: jsonschema.ObjectType,
}, nil
}
// WE ERROR OUT IF THERE ARE CYCLES IN THE JSON SCHEMA