Apache Sling GraphQL Schema Aggregator

Clone this repo:
  1. 02f7843 SLING-10676 - remove SECURITY.md which is not needed by Bertrand Delacretaz · 3 days ago master
  2. b103474 SLING-10681 - Restrict the naming pattern for partials by Radu Cotescu · 4 days ago
  3. 1d39951 SLING-10676 - add or update SECURITY.md by Bertrand Delacretaz · 5 days ago
  4. 0729894 SLING-10676 - add or update SECURITY.md by Bertrand Delacretaz · 5 days ago
  5. cb36834 SLING-10676 - add or update SECURITY.md by Bertrand Delacretaz · 6 days ago

Apache Sling

Build Status Test Status Coverage Sonarcloud Status graphql License

Apache Sling GraphQL Schema Aggregator

This module is one of several which provide GraphQL support for Apache Sling.

The Sling GraphQL Schema Aggregator provides services to combine partial GraphQL schema (“partials”) supplied by provider bundles.

The partials are structured text files, supplied as OSGi bundle resources, that provide sections (like query, mutation, types sections) that are aggregated to build a GraphQL Schema using the SDL (Schema Definition Language) syntax.

A GraphQL schema must contain one Query statement and can contain a most one Mutation statement, so partials cannot be assembled by just concatenating them. The schema assembler defines a simple section-based syntax for the partials, so that they can be aggregated efficiently.

This module also provides a SchemaAggregatorServlet that generates schemas by aggregating partials, by mapping request selectors to lists of partial names. The result can be used directly by the Sling GraphQL Core module, which makes an internal Sling request to get the schema.

Partials can also depend on others by declaring the required dependencies by name, to make sure the aggregated schemas are valid.

With this mechanism, an OSGi bundle can provide both a partial schema and the Sling data fetching and processing services that go with it. This allows a GraphQL “API plane” (usually defined by a specific instance of the Sling GraphQLServlet) to be built out of several OSGi bundles which each focus on a specific set of queries, mutations and types.

Provider bundles

To provide partials, a bundle sets a Sling-GraphQL-Schema header in its OSGi manifest, with a value that points to a path under which partials are found in the bundle resources.

A provider bundle also needs to declare the following requirement, to allow the schema aggregator to receive its bundle start event:


The version number range must be consistent with this module's version.

A partial is a text file with the structure described below. As usual, The Truth Is In The Tests, see the example partial in the test sources for a reference that's guaranteed to be valid.

# Example GraphQL partial schema
# Any text before the first section is ignored.

PARTIAL: Example GraphQL schema partial
The contents of the PARTIAL section are ignored, only its
description (the text follows the PARTIAL section name
above) is used.

PARTIAL is the only required section.

REQUIRE: base.scalars, base.schema
The description of the optional REQUIRE section is a
comma-separated list of partials which are required for this
one to be valid. The content of this section is ignored, only
its description is used to build that list.

The content of the optional PROLOGUE section is concatenated
in the aggregated schema, before all the other sections.

The content of the optional QUERY sections of all partials
is aggregated in a `type QUERY {...}` section in the output.

Like for the QUERY section, the content of the optional
MUTATION sections of all partials is aggregated in
a `type MUTATION {...}` section in the output.

The content of the TYPES sections of all partials is
aggregated in the output, after all the other sections.

Partial names

The name of a partial, used in the selector mappings of the SchemaAggregatorServlet, is defined by its filename in the bundle resources, omitting the file extension. A partial found under /path-set-by-the-bundle-header/this.is.txt in its bundle is named this.is . Partial names must be unique system-wide, so it's good to use some form of namespacing or agreed upon naming convention for them.

SchemaAggregatorServlet configuration

Here's a configuration example from the test code.

// Configure the org.apache.sling.graphql.schema.aggregator.SchemaAggregatorServlet
    .put("sling.servlet.resourceTypes", "sling/servlet/default")

    // The extension must be the one used by the GraphQLServlet to retrieve schemas
    // which by default is 'GQLschema'
    .put("sling.servlet.extensions", GQL_SCHEMA_EXT)

    // The GraphQLServlet uses an internal GET request for the schema
    .put("sling.servlet.methods", new String[] { "GET" })

    // Several selectors can be configured to setup API planes, each with their own GraphQL schema
    .put("sling.servlet.selectors", new String[] { "X", "Y" })

    // This mapping defines which partials to use to build the schema for each selector
    // The lists can use either the exact names of partials, or (Java flavored) regular expressions on
    // their names, identified by a starting and ending slash.
    .put("selectors.to.partials.mapping", new String[] { "X:firstA,secondB", "Y:secondA,firstB,/second.*/" })

TODO / wishlist

Invalid section names in partials should cause parsing errors.

The REQUIRES section of partial should be translated to OSGi capabilities, to be able to detect missing requirements at system assembly time or using the Feature Model Analyser.

We'll probably need a utility to aggregate schemas for automated tests, to allow test code to include required schema partials.

Errors like invalid or missing partials are currently only logged, it would be useful to have them cause louder errors, like schema aggregation failing with error messages when things went wrong, and/or this module providing a Health Check service to detect problems.

Caching is probably not needed in this module, as the GraphQL Core caches compiled schemas.