Skip to content

fix: skip validation of secondary instances without context

Administrator requested to merge fix/skip_validation_of_secondary_instances_without_context into main

Created by: ashleythedeveloper

This PR fixes #313 (closed) and also addresses a secondary issue that was hidden by the primary problem.

Primary Issue: Validation Pipeline Validating Too Much

Context

As described in #313 (closed), the validation pipeline was incorrectly validating all instances in the payload, not just the primary sample instance. Secondary instances lacking a @context caused the pipeline to fail when the JSON-LD processor attempted to expand them.

Fix

I've introduced a JARGON_INSTANCES_TO_VALIDATE variable that explicitly lists which sample instance names should be validated.

The code now filters the payload to validate only those named instances, ignoring all others within the JSON-LD expansion step.

This approach was chosen because:

  • The instance order is not guaranteed.
  • There is no identifier marking the “primary” instance.
  • The primary instance’s name is user-defined and not derived from the domain name.

While I considered making this configurable via an environment variable, I decided to keep it under version control to avoid accidental or unwanted changes.

Secondary Issue: Unable to Fetch Context

Context

Even after isolating the primary instance, the JSON-LD validation step failed because it couldn’t fetch the remote UNTP context.

This was due to the context file not being published at the specified IRI:

  • IRI is set to https://test.uncefact.org/vocabulary/untp/{{CREDENTIAL_ACRONYM}}/{{version}}/ as per domain settings in Jargon.
  • However, per our updated release workflow, we no longer release directly to test.uncefact.org/vocabulary/untp/, we now publish the context files to this repository, validate the artifacts in the pipeline, and then publish to test.uncefact.org/vocabulary/untp/.

Fix

I've introduced a JARGON_CONTEXT_IRI_PREFIX variable set to https://test.uncefact.org/vocabulary/untp/ to detect and replace the IRI.

The code now reads the @context array and rewrites any IRI matching that prefix to use the Jargon-hosted context file provided in the payload.

The pipeline works as expected and has been tested with the available payloads from previous failed workflows.

Merge request reports