Gradle’s dependency management engine is variant aware.
In the previous section, Gradle built a graph of resolved dependencies. During graph resolution, Gradle selected the proper variants of each dependency based on the requirements of the build.
Variants represent different ways a component can be used, such as for Java compilation, native linking, or documentation. Each variant may have its own artifacts and dependencies.
Gradle uses attributes to determine which variant to choose. These attributes add context to each variant, describing when they should be used.
ComponentsLet’s review our description of components in the previous section. A component:
contains variants
which contain one or more artifacts
which contain zero or more dependencies
which is described by metadata
In the example above, org.jetbrains.kotlinx:kotlinx-serialization-json:1.5.1
is the component. It has a module org.jetbrains.kotlinx:kotlinx-serialization-json
and a version 1.5.1
:
Variants represent different versions or aspects of a component, like api
vs implementation
or jar
vs classes
.
In the example above, org.jetbrains.kotlinx:kotlinx-serialization-json:1.5.1
has four variants: jvm
, android
, js
, and native
.
Attributes are type-safe key-value pairs used by both the consumer and the producer during variant selection: attribute : value
.
In the example above, there are two attributes of importance for the variants of org.jetbrains.kotlinx:kotlinx-serialization-json:1.5.1
:
JVM
org.gradle.usage: java-runtime
org.jetbrains.kotlin.platform.type: jvm
Android
org.gradle.usage: java-runtime
org.jetbrains.kotlin.platform.type: androidJvm
Javascript
org.gradle.usage: java-runtime
org.jetbrains.kotlin.platform.type: js
Native
org.gradle.usage: native-link
org.jetbrains.kotlin.platform.type: native
Note that the metadata of org.jetbrains.kotlinx:kotlinx-serialization-json:1.5.1
showcases many more attributes to describe its variants such as org.gradle.libraryelements
or org.gradle.category
. However, Gradle may use any number of them during dependency resolution as some attributes are more relevant than others:
"variants": [
...
{
"name": "jsLegacyRuntimeElements-published",
"attributes": {
"org.gradle.category": "library",
"org.gradle.usage": "kotlin-runtime",
"org.jetbrains.kotlin.js.compiler": "legacy",
"org.jetbrains.kotlin.platform.type": "js"
}
},
{
"name": "jvmRuntimeElements-published",
"attributes": {
"org.gradle.category": "library",
"org.gradle.libraryelements": "jar",
"org.gradle.usage": "java-runtime",
"org.jetbrains.kotlin.platform.type": "jvm"
}
},
...
]
There are two types of attributes Gradle uses to match the available variants to the one required for the build:
Consumer attributes: Define the desired characteristics of a variant requested by a resolvable configuration.
Producer attributes: Each variant has a set of attributes that describe its purpose.
There are no restrictions on how many variants a component can define.
A typical component will include at least an implementation variant but may also provide additional variants, such as test fixtures, documentation, or source code. Furthermore, a component can offer different variants for the same usage, depending on the consumer. For instance, during compilation, a component may provide different headers for Linux, Windows, and macOS.
Gradle performs variant-aware selection by matching the attributes specified by the consumer with those defined by the producer using an attribute matching algorithm.
The variant name is primarily used for debugging and error messages. It does not play a role in variant matching; only the variant’s attributes are used in the matching process. Attribute Matching AlgorithmGradle’s dependency resolution engine follows this attribute matching algorithm to determine the best candidate (or fail if no match is found):
Step 1: Find Compatible CandidatesThe requested attributes are matched against the candidate attributes.
FOR each candidate:
FOR each candidate attribute:
IF the candidate attribute is missing, mark the candidate as missing.
ELSE IF the candidate attribute exactly matches the requested attribute, the candidate remains compatible.
ELSE IF the candidate attribute passes the compatibility test, the candidate remains compatible.
ELSE, the candidate is eliminated.
After filtering, Gradle checks if one candidate is a strict superset of all others.
IF only one candidate remains, that candidate wins.
IF multiple candidates remain, check if one:
Matches more attributes than any other.
Has no missing attributes compared to another candidate.
IF such a strict superset exists, that candidate wins.
If multiple candidates remain, Gradle attempts disambiguation:
Requested attributes are sorted by precedence.
FOR each requested attribute in precedence order:
IF a disambiguation rule exists, apply it.
IF a requested attribute has the same value across all remaining candidates, ignore it.
IF only one candidate remains, that candidate wins.
If multiple candidates still remain, Gradle processes unordered attributes (those not in the precedence list):
FOR each requested attribute without precedence order:
IF a disambiguation rule exists, apply it.
IF a requested attribute has the same value across all remaining candidates, ignore it.
IF only one candidate remains, that candidate wins.
IF no candidates remain, Gradle falls back to the most compatible candidates (step 2).
If multiple candidates still remain, Gradle considers extra attributes:
These are attributes that exist only in some candidates.
Candidates that lack extra attributes are preferred.
The attribute precedence order is considered.
First, Gradle processes prioritized extra attributes:
FOR each extra attribute in precedence order:
IF some candidates have the attribute and others donât:
Candidates that lack the extra attribute are preferred.
Candidates with the extra attribute are eliminated.
IF only one candidate remains, that candidate wins.
IF no candidates remain, Gradle falls back to the original compatible candidates (step 1).
Then, Gradle processes unordered extra attributes:
FOR each remaining extra attribute without defined precedence:
IF the attribute is present in only some candidates:
Candidates without the attribute are preferred.
Candidates with the extra attribute are eliminated.
IF only one candidate remains, that candidate wins.
IF no candidates remain, Gradle falls back to the original compatible candidates (step 1).
If candidates still remain, Gradle eliminates candidates based on requested attributes:
FOR each extra attribute:
IF all candidates provide the same value, ignore it.
ELSE, remove candidates that do provide it.
IF only one candidate remains, that candidate wins.
Unless otherwise indicated, if at any point no candidates remain, resolution fails.
Additionally, Gradle outputs a list of all compatible candidates from step 1 to help with debugging attribute matching failures.
Plugins and ecosystems can influence the selection algorithm by implementing compatibility rules, disambiguation rules, and defining the precedence of attributes. Attributes with a higher precedence are used to eliminate candidates in order.
For example, in the Java ecosystem, the org.gradle.usage
attribute has a higher precedence than org.gradle.libraryelements
. This means that if two candidates were available with compatible values for both org.gradle.usage
and org.gradle.libraryelements
, Gradle will choose the candidate that passes the disambiguation rule for org.gradle.usage
.
There are two exceptions to the variant-aware resolution process:
When a producer has no variants, a default artifact is selected.
When a consumer explicitly selects a configuration by name, the artifacts associated with that configuration are used.
Letâs walk through an example where a consumer is trying to use a library for compilation.
First, the consumer details how it’s going to use the result of dependency resolution. This is achieved by setting attributes on the consumer’s resolvable configuration.
In this case, the consumer wants to resolve a variant that matches org.gradle.usage=java-api
.
Next, the producer exposes different variants of its component:
API variant (named apiElements
) with the attribute org.gradle.usage=java-api
Runtime variant (named runtimeElements
) with the attribute org.gradle.usage=java-runtime
Finally, Gradle evaluates the variants and selects the correct one:
The consumer requests a variant with attributes org.gradle.usage=java-api
The producerâs apiElements
variant matches this request.
The producerâs runtimeElements
variant does not match.
org.gradle.usage=java-api
org.gradle.usage=java-api
apiElements
â Yes
org.gradle.usage=java-api
org.gradle.usage=java-runtime
runtimeElements
â No
As a result, Gradle selects the apiElements
variant and provides its artifacts and dependencies to the consumer.
In real-world scenarios, both consumers and producers often work with multiple attributes.
For instance, a Java Library project in Gradle will involve several attributes:
Attribute Descriptionorg.gradle.usage
Describes how the variant is used.
org.gradle.dependency.bundling
Describes how the variant handles dependencies (e.g., shadow jar, fat jar, regular jar).
org.gradle.libraryelements
Describes the packaging of the variant (e.g., classes or jar).
org.gradle.jvm.version
Describes the minimal version of Java the variant targets.
org.gradle.jvm.environment
Describes the type of JVM the variant targets.
Letâs consider a scenario where the consumer wants to run tests using a library on Java 8, and the producer supports two versions: Java 8 and Java 11.
Step 1: Consumer specifies the requirements.
The consumer wants to resolve a variant that:
Can be used at runtime (org.gradle.usage=java-runtime
).
Can run on at least Java 8 (org.gradle.jvm.version=8
).
Step 2: Producer exposes multiple variants.
The producer offers variants for both Java 8 and Java 11 for both API and runtime usage:
Variant Name AttributesapiJava8Elements
org.gradle.usage=java-api
, org.gradle.jvm.version=8
runtime8Elements
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
apiJava11Elements
org.gradle.usage=java-api
, org.gradle.jvm.version=11
runtime11Elements
org.gradle.usage=java-runtime
, org.gradle.jvm.version=11
Step 3: Gradle matches the attributes.
Gradle compares the consumer's requested attributes with the producer's variants:
The consumer requests a variant with org.gradle.usage=java-runtime
and org.gradle.jvm.version=8
.
Both runtime8Elements
and runtime11Elements
match the org.gradle.usage=java-runtime
attribute.
The API variants (apiJava8Elements
and apiJava11Elements
) are discarded as they don’t match org.gradle.usage=java-runtime
.
The variant runtime8Elements
is selected because it is compatible with Java 8.
The variant runtime11Elements
is incompatible because it requires Java 11.
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
runtime8Elements
â Selected
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
org.gradle.usage=java-runtime
, org.gradle.jvm.version=11
runtime11Elements
â Incompatible
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
org.gradle.usage=java-api
, org.gradle.jvm.version=8
apiJava8Elements
â Discarded
org.gradle.usage=java-runtime
, org.gradle.jvm.version=8
org.gradle.usage=java-api
, org.gradle.jvm.version=11
apiJava11Elements
â Discarded
Gradle selects runtime8Elements
and provides its artifacts and dependencies to the consumer.
But what happens if the consumer sets org.gradle.jvm.version=7
?
In this case, dependency resolution would fail, with an error explaining there is no suitable variant. Gradle knows the consumer requires a Java 7-compatible library, but the producer's minimum version is 8.
If the consumer requested org.gradle.jvm.version=15
, Gradle could choose either the Java 8 or Java 11 variant. Gradle would then select the highest compatible versionâJava 11.
Gradle offers built-in tasks to visualize the variant selection process and display the producer and consumer attributes involved.
Outgoing variants reportThe report task outgoingVariants
shows the list of variants available for selection by consumers of the project. It displays the capabilities, attributes and artifacts for each variant.
This task is similar to the dependencyInsight
reporting task.
By default, outgoingVariants
prints information about all variants. It offers the optional parameter --variant <variantName>
to select a single variant to display. It also accepts the --all
flag to include information about legacy and deprecated configurations, or --no-all
to exclude this information.
Here is the output of the outgoingVariants
task on a freshly generated java-library
project:
> Task :outgoingVariants -------------------------------------------------- Variant apiElements -------------------------------------------------- API elements for the 'main' feature. Capabilities - new-java-library:lib:unspecified (default capability) Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.version = 11 - org.gradle.libraryelements = jar - org.gradle.usage = java-api Artifacts - build/libs/lib.jar (artifactType = jar) Secondary Variants (*) -------------------------------------------------- Secondary Variant classes -------------------------------------------------- Description = Directories containing compiled class files for main. Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.version = 11 - org.gradle.libraryelements = classes - org.gradle.usage = java-api Artifacts - build/classes/java/main (artifactType = java-classes-directory) -------------------------------------------------- Variant mainSourceElements (i) -------------------------------------------------- Description = List of source directories contained in the Main SourceSet. Capabilities - new-java-library:lib:unspecified (default capability) Attributes - org.gradle.category = verification - org.gradle.dependency.bundling = external - org.gradle.verificationtype = main-sources Artifacts - src/main/java (artifactType = directory) - src/main/resources (artifactType = directory) -------------------------------------------------- Variant runtimeElements -------------------------------------------------- Runtime elements for the 'main' feature. Capabilities - new-java-library:lib:unspecified (default capability) Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.version = 11 - org.gradle.libraryelements = jar - org.gradle.usage = java-runtime Artifacts - build/libs/lib.jar (artifactType = jar) Secondary Variants (*) -------------------------------------------------- Secondary Variant classes -------------------------------------------------- Description = Directories containing compiled class files for main. Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.version = 11 - org.gradle.libraryelements = classes - org.gradle.usage = java-runtime Artifacts - build/classes/java/main (artifactType = java-classes-directory) -------------------------------------------------- Secondary Variant resources -------------------------------------------------- Description = Directories containing the project's assembled resource files for use at runtime. Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.version = 11 - org.gradle.libraryelements = resources - org.gradle.usage = java-runtime Artifacts - build/resources/main (artifactType = java-resources-directory)
From this you can see the two main variants that are exposed by a java library, apiElements
and runtimeElements
. Notice that the main difference is on the org.gradle.usage
attribute, with values java-api
and java-runtime
. As they indicate, this is where the difference is made between what needs to be on the compile classpath of consumers, versus what’s needed on the runtime classpath.
It also shows secondary variants, which are exclusive to Gradle projects and not published. For example, the secondary variant classes
from apiElements
is what allows Gradle to skip the JAR creation when compiling against a java-library
project.
A project cannot have multiple configurations with the same attributes and capabilities. In that case, the project will fail to build.
In order to be able to visualize such issues, the outgoing variant reports handle those errors in a lenient fashion. This allows the report to display information about the issue.
Resolvable configurations reportGradle also offers a complimentary report task called resolvableConfigurations
that displays the resolvable configurations of a project, which are those which can have dependencies added and be resolved.
The report will list their attributes and any configurations that they extend. It will also list a summary of any attributes which will be affected by Compatibility Rules or Disambiguation Rules during resolution.
By default, resolvableConfigurations
prints information about all purely resolvable configurations. These are configurations that are marked resolvable but not marked consumable. Though some resolvable configurations are also marked consumable, these are legacy configurations that should not have dependencies added in build scripts.
This report offers:
An optional parameter:
--configuration <configurationName>
â Selects a single configuration to display.
Flags for including or excluding legacy and deprecated configurations:
--all
â Includes information about legacy and deprecated configurations.
--no-all
â Excludes this information.
Flags for controlling transitive extensions in the extended configurations section:
--recursive
â Lists configurations that are extended transitively rather than directly.
--no-recursive
â Excludes this information.
Here is the output of the resolvableConfigurations
task on a freshly generated java-library
project:
> Task :resolvableConfigurations -------------------------------------------------- Configuration compileClasspath -------------------------------------------------- Description = Compile classpath for source set 'main'. Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.environment = standard-jvm - org.gradle.jvm.version = 11 - org.gradle.libraryelements = classes - org.gradle.usage = java-api Extended Configurations - compileOnly - implementation -------------------------------------------------- Configuration runtimeClasspath -------------------------------------------------- Description = Runtime classpath of source set 'main'. Attributes - org.gradle.category = library - org.gradle.dependency.bundling = external - org.gradle.jvm.environment = standard-jvm - org.gradle.jvm.version = 11 - org.gradle.libraryelements = jar - org.gradle.usage = java-runtime Extended Configurations - implementation - runtimeOnly [...]
From this you can see the two main configurations used to resolve dependencies, compileClasspath
and runtimeClasspath
, as well as their corresponding test configurations (truncated).
Now that we understand variant selection and attribute matching, let’s move on to the artifact resolution phase of dependency resolution. This phase is also variant aware.
Next Step: Learn about Artifact Resolution >>
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4