def toConfig1 = configuration(toComponent, "to-1")
        def toConfig2 = configuration(toComponent, "to-2")
        fromConfig.hierarchy >> ImmutableSet.of("from")
        fromConfig2.hierarchy >> ImmutableSet.of("child", "from")
        fromConfig3.hierarchy >> ImmutableSet.of("other")

        Multimap<String, String>  configMapping = LinkedHashMultimap.create()
        configMapping.put("*", "to-2")

     *
     * @see #getModuleNames()
     */
    private final Map<Path, String> descriptors;

    /**
     * Whether module hierarchy was detected. If false, then package hierarchy is assumed.
     * In a package hierarchy, the {@linkplain #descriptors} map has either zero or one entry.
     * In a module hierarchy, the descriptors map may have an arbitrary number of entries,
     * including one (so the map size cannot be used as a criterion).
     *

        immutable.getConfiguration("compile").hierarchy as List == ["compile"]
        immutable.getConfiguration("runtime").hierarchy as List == ["runtime", "compile"]
        immutable.getConfiguration("master").hierarchy as List == ["master"]
        immutable.getConfiguration("test").hierarchy as List == ["test", "runtime", "compile"]
        immutable.getConfiguration("default").hierarchy as List == ["default", "runtime", "compile", "master"]

            }
        });

        // Collect all dependencies and excludes in hierarchy.
        // After running the dependency actions and preventing from mutation above, we know the
        // hierarchy will not change anymore and all configurations in the hierarchy
        // will no longer be mutated.
        ImmutableSet<String> hierarchy = Configurations.getNames(configuration.getHierarchy());

            return null;
        }

        ImmutableSet<String> hierarchy = constructHierarchy(descriptorConfiguration);
        boolean transitive = descriptorConfiguration.isTransitive();
        boolean visible = descriptorConfiguration.isVisible();
        return createConfiguration(getId(), name, transitive, visible, hierarchy, variantMetadataRules);
    }

1. We generate interceptors from the annotations declared in the source code 
2. We collect Gradle API type hierarchy

The first one is important for both configuration cache instrumentation and API upgrades while the second one is used just for API upgrades. 
Collecting the Gradle API type hierarchy is fully automatic, while declaring interceptors have to be done by Gradle developers. 

```mermaid
flowchart TB

    public DefaultSerializerRegistry() {
        this(true);
    }

    public DefaultSerializerRegistry(boolean supportClassHierarchy) {
        this.classMatcher = supportClassHierarchy ? SerializerClassMatcherStrategy.HIERARCHY : SerializerClassMatcherStrategy.STRICT;
    }

    @Override
    public <T> void register(Class<T> implementationType, final Serializer<T> serializer) {

    private final OutputEventListener listener;
    private final LogHeaderFormatter headerFormatter;
    private final boolean verbose;

    // Maintain a hierarchy of all progress operations in progress — heads up: this is a *forest*, not just 1 tree
    private final Map<OperationIdentifier, OperationState> operationsInProgress = new LinkedHashMap<OperationIdentifier, OperationState>();

import org.gradle.vcs.SourceControl;

import javax.annotation.Nullable;
import java.io.File;
import java.util.Arrays;

/**
 * <p>Declares the configuration required to instantiate and configure the hierarchy of {@link
 * org.gradle.api.Project} instances which are to participate in a build.</p>
 *
 * <p>There is a one-to-one correspondence between a <code>Settings</code> instance and a <code>{@value

     *
     * A path is contained in the set if it or one of its ancestors has
     * been added to the set.
     */
    public abstract boolean contains(String path);

    /**
     * Whether this hierarchy is empty, i.e. contains no directories.
     */
    public abstract boolean isEmpty();

    /**