After constructing a dependency graph, Gradle can perform artifact resolution on the resolved graph.

Gradle APIs can be used to influence the process of artifact selection — the mapping of a graph to a set of artifacts.

Gradle can then expose the results of artifact selection as an ArtifactCollection. More commonly, the results are exposed as a FileCollection, which is a flat list of files.

Artifact selection

Artifact selection operates on the dependency graph on a node-by-node basis. Each node in the graph may expose multiple sets of artifacts, but only one of those sets may be selected. For example, the runtimeElements variant of the Java plugins exposes a jar, classes, and resources artifact set. These three artifact sets represent the same distributable, but in different forms.

For each node (variant) in a graph, Gradle performs attribute matching over each set of artifacts exposed by that node to determine the best artifact set. If no artifact sets match the requested attributes, Gradle will attempt to construct an artifact transform chain to satisfy the request.

For more details on the attribute matching process, see the attribute matching section.

Implicit artifact selection

By default, the attributes used for artifact selection are the same as those used for variant selection during graph resolution. These attributes are specified by the Configuration#getAttributes() property.

To perform artifact selection (and implicitly, graph resolution) using these default attributes, use the FileCollection and ArtifactCollection APIs.

Files can also be accessed from the configuration’s ResolvedConfiguration, LenientConfiguration, ResolvedArtifact and ResolvedDependency APIs. However, these APIs are in maintenance mode and are discouraged for use in new development. These APIs perform artifact selection using the default attributes.

Resolving files

To resolve files, we first define a task that accepts a ConfigurableFileCollection as input:

KotlinGroovy
build.gradle
abstract class ResolveFiles extends DefaultTask {

    @InputFiles
    abstract ConfigurableFileCollection getFiles()

    @TaskAction
    void print() {
        files.each {
            println(it.name)
        }
    }
}

Then, we can wire up a resolvable configuration’s files to the task’s input. The Configuration directly implements FileCollection and can be wired directly. Alternatively, wiring through Configuration#getIncoming() is a more explicit approach:

KotlinGroovy
build.gradle
tasks.register("resolveConfiguration", ResolveFiles) {
    files.from(configurations.runtimeClasspath)
}
tasks.register("resolveIncomingFiles", ResolveFiles) {
    files.from(configurations.runtimeClasspath.incoming.files)
}

Running both of these tasks, we can see the output is identical:

> Task :resolveConfiguration
junit-platform-commons-1.11.0.jar
junit-jupiter-api-5.11.0.jar
opentest4j-1.3.0.jar

> Task :resolveIncomingFiles
junit-platform-commons-1.11.0.jar
junit-jupiter-api-5.11.0.jar
opentest4j-1.3.0.jar

Resolving artifacts

Instead of consuming the files directly from the implicit artifact selection process, we can consume the artifacts, which contain both the files and the metadata.

This process is slightly more complicated, as in order to maintain Configuration Cache compatibility, we need to split the fields of ResolvedArtifactResult into two task inputs:

KotlinGroovy
build.gradle
class ArtifactDetails {
    ComponentArtifactIdentifier id
    ResolvedVariantResult variant

    ArtifactDetails(ComponentArtifactIdentifier id, ResolvedVariantResult variant) {
        this.id = id
        this.variant = variant
    }
}

abstract class ResolveArtifacts extends DefaultTask {

    @Input
    abstract ListProperty<ArtifactDetails> getDetails()

    @InputFiles
    abstract ListProperty<File> getFiles()

    void from(Provider<Set<ResolvedArtifactResult>> artifacts) {
        details.set(artifacts.map {
            it.collect { artifact -> new ArtifactDetails(artifact.id, artifact.variant) }
        })
        files.set(artifacts.map {
            it.collect { artifact -> artifact.file }
        })
    }

    @TaskAction
    void print() {
        List<ArtifactDetails> allDetails = details.get()
        List<File> allFiles = files.get()

        assert allDetails.size() == allFiles.size()
        for (int i = 0; i < allDetails.size(); i++) {
            def details = allDetails.get(i)
            def file = allFiles.get(i)
            println("${details.variant.displayName}:${file.name}")
        }
    }
}

This task is initialized similarly to the file resolution task:

KotlinGroovy
build.gradle
tasks.register("resolveIncomingArtifacts", ResolveArtifacts) {
    from(configurations.runtimeClasspath.incoming.artifacts.resolvedArtifacts)
}

Running this task, we can see that file metadata is included in the output:

org.junit.platform:junit-platform-commons:1.11.0 variant runtimeElements:junit-platform-commons-1.11.0.jar
org.junit.jupiter:junit-jupiter-api:5.11.0 variant runtimeElements:junit-jupiter-api-5.11.0.jar
org.opentest4j:opentest4j:1.3.0 variant runtimeElements:opentest4j-1.3.0.jar

Customizing artifact selection

An ArtifactView operates on top of the resolved dependency graph (i.e., ResolutionResult) but allows you to apply different attributes.

When you call a configuration’s getFiles(), Gradle selects artifacts based on the attributes used during graph resolution. However, an ArtifactView is more flexible. It allows you to resolve artifacts from the graph with custom attributes.

An ArtifactView allows you to:

  1. Query artifacts with different attributes:

    • Suppose the graph resolved a dependency’s runtime variant. You can use an ArtifactView to extract artifacts from its api variant instead, even if they weren’t originally part of the resolved graph.

  2. Extract specific types of artifacts:

    • You can request only the .jar files or a specific artifact type (e.g., sources, Javadoc) by specifying an attribute like artifactType.

  3. Avoid side effects:

    • Using an ArtifactView allows you to extract artifacts without changing the underlying dependency resolution logic or configuration state.

In the following example, a producer project creates a library with the following variants and their attributes: