[
https://issues.apache.org/jira/browse/SLING-10619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17385358#comment-17385358
]
Bertrand Delacretaz commented on SLING-10619:
---------------------------------------------
bq. ...they also in turn advertise a capability for each schema / partial they
provide...
Do you envision generating those capabilities automatically at build time?
Maybe with a bnd plugin like
https://github.com/apache/sling-org-apache-sling-bnd-models ?
I think from the developer's point of view the ideal scenario should be:
* Add the Sling-GraphQL-Schema header pointing to the path where partials are
found
* Add any partial schema files in there
* That's it
And at build time the tooling should:
* a) Validate the syntax of the partial files
* b) Generate the corresponding OSGi provide-capability and require-capability
headers
And maybe later a Sling feature analyser can
* c) verify that the required partials are provided as well as ideally verify
the syntax of the aggregated schema.
That's the ideal case but I think we need at least b) to avoid having to
duplicate information, and that requires a) to extract the relevant values.
> Remove sling.graphql-schema-aggregator capability if not useful
> ---------------------------------------------------------------
>
> Key: SLING-10619
> URL: https://issues.apache.org/jira/browse/SLING-10619
> Project: Sling
> Issue Type: Improvement
> Components: GraphQL
> Reporter: Bertrand Delacretaz
> Priority: Minor
> Fix For: GraphQL Schema Aggregator 0.0.2
>
>
> The {{ProviderBundleTracker}} currently requires provider bundles to define
> an [OSGi
> capability|https://github.com/apache/sling-org-apache-sling-graphql-schema-aggregator/blob/04474a2fe16389c0df60471922f38e7dcfc637ef/src/main/java/org/apache/sling/graphql/schema/aggregator/impl/ProviderBundleTracker.java#L48],
> I think the idea is for that class to efficiently ignore bundles which are
> not provider bundles.
> However I don't think this is more efficient than just having
> {{addingBundle(...)}} test for the {{Sling-GraphQL-Schema}} header.
> I would argue that doing the following in that method would be at least as
> efficient than the current code in terms of ignoring non-provider bundles:
> {code:java}
> @Override
> public Object addingBundle(Bundle bundle, BundleEvent event) {
> final String providersPath =
> bundle.getHeaders().get("Sling-GraphQL-Schema");
> if (providersPath == null) {
> // not interested in that bundle
> return;
> ...
> {code}
> Currently, requiring this capability means provider bundles must declare both
> the {{Sling-GraphQL-Schema}} header and the
> {{sling.graphql-schema-aggregator}} capability.
> Unless there are definite advantages in having both, I'd be in favor of
> requiring just the {{Sling-GraphQL-Schema}} header to keep things as simple
> as possible.
> [~radu] I know you were in favor of using the capability, if we agree that
> the above code has no performance impact, do you see another benefit of using
> the capability?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
