JApicmp Gradle Plugin

The japicmp-gradle-plugin provides binary compatibility reporting through JApicmp using Gradle.

Installation

Note	Version 0.2+ of this plugin is NOT compatible with 0.1

This plugin requires Gradle 4+. Use the following snippet inside a Gradle build file:

build.gradle

buildscript {
    repositories {
        jcenter()
    }

    dependencies {
        classpath 'me.champeau.gradle:japicmp-gradle-plugin:{plugin-version}'
    }
}
apply plugin: 'me.champeau.gradle.japicmp'

plugins {
    id 'me.champeau.gradle.japicmp' version '{plugin-version}'
}

Configuration

The plugin provides a new task type: me.champeau.gradle.japicmp.JapicmpTask that you can use to compare two jars. This task exposes the following properties as part of its configuration:

oldClasspath	The classpath of the baseline library to compare. Type: FileCollection
newClasspath	The classpath of the current version of the library, which you want to check binary compatibility Type: FileCollection
oldArchives	The jar files which will be used as the baseline for comparison. Type: FileCollection.
newArchives	The jar files we want to analyze. Type: Type: FileCollection.
onlyModified	Outputs only modified classes/methods. If not set to true, all classes and methods are printed. Type: boolean. Default value: false
onlyBinaryIncompatibleModified	Outputs only classes/methods with modifications that result in binary incompatibility. Type: boolean. Default value: false
packagesToInclude	List of package names to include, * can be used as wildcard. Type: List<String>
packagesToExclude	List of package names to exclude, * can be used as wildcard. Type: List<String>
accessModifier	Sets the access modifier level (public, package, protected, private). Type: String. Default value: public
failOnModification	When set to true, the build fails in case a modification has been detected. Type: boolean. Default value: false
xmlOutputFile	Path to the generated XML report. Type: File. Default value: null
htmlOutputFile	Path to the generated HTML report. Type: File. Default value: null
txtOutputFile	Path to the generated TXT report. Type: File. Default value: null
includeSynthetic	Synthetic classes and class members (like e.g. bridge methods) are not tracked per default. This new option enables the tracking of such kind of classes and class members

If you don’t set oldArchives and newArchives, the plugin will infer them from the oldClasspath and newClasspath properties: * if you set the classpath to a configuration, the archives to compare will be the first level dependencies of that configuration * if you set the classpath to a simple file collection, all archives will be compared

Usage

Add the following to your build file:

task japicmp(type: me.champeau.gradle.japicmp.JapicmpTask) {
    oldClasspath = files('path/to/reference.jar')
    newClasspath = files(jar.archivePath')
    onlyModified = true
    failOnModification = true
    txtOutputFile = file("$buildDir/reports/japi.txt")
}

Custom reports and failure conditions

The plugin supports a DSL to generate custom reports based on the API comparison result. This has several advantages:

you can generate a report that focuses only on your public API, leaving the internal APIs out
you can implement custom rules to determine if the build should fail or not
the report can be presented to users and provide guidance for migration from one version to the other

Configuration

The report can be configured using the richReport block:

task japicmp(type: me.champeau.gradle.japicmp.JapicmpTask) {
   ...
   richReport {
      ...
   }
}

Options for the rich report are:

renderer	The renderer used to generate the report. By default, it uses the GroovyReportRenderer
includedClasses	A list of strings representing inclusion patterns (interpreted as regular expressions). Only classes matching this pattern will be included.
excludedClasses	A list of strings representing exclusion patterns. If a class fully qualified name matches any of those patterns, it will not be included.
destinationDir	the directory where to store the report
reportName	file name of the generated report (defaults to `rich-report.html`)
title	a title for the report
description	a description for the report

Custom rules

Rules are used to add violations to the report. The "violation" term must be taken in a simple sense, as it represents data to be shown in the report, whether it’s a critical violation or just information.

A violation consists of a triplet (member, severity, explanation), that will be seen in the report. For example, if a binary incompatibility is found, you can create a violation using:

Violation.notBinaryCompatible(member)

which will automatically assign it to the error severity, leading in a build failure. However, it is possible to create any kind of violation, and even accept binary incompatible changes.

Rules can be applied to 3 different levels:

all members (a generic rule applied unconditionnaly)
on specific change types (NEW, REMOVED, UNCHANGED, MODIFIED), see JApiChangeStatus
on specific compatibility change descriptors (see JApiCompatibilityChange)

For example, imagine that we want to check that all new methods are annotated with @Incubating (this is a rule in the Gradle project). Then, you need to create a rule class which will implement that check:

class IncubatingMissingRule implements ViolationRule {
    @Override
    Violation maybeViolation(final JApiCompatibility member) {
        if (member instanceof JApiMethod) {
            if (!member.annotations.find { it.fullyQualifiedName == 'org.gradle.api.Incubating' }) {
                if (!member.jApiClass.annotations.find {
                    it.fullyQualifiedName == 'org.gradle.api.Incubating'
                }) {
                    Violation.error(member, "New method is not annotated with @Incubating")
                }
            }
        }
    }
}

and then you need to configure the report to use that rule:

richReport {
   addRule(JApiChangeStatus.NEW, IncubatingMissingRule)
}

Rules can take arguments, but those are limited to Map<String, String>. For example, the following rule will mark a binary breaking change as an error, unless it is reviewed and accepted. The list of acceptations is passed as an argument to the rule:

class AcceptedRegressionRule implements ViolationRule {
    private final Map<String, String> acceptedViolations

    public AcceptedRegressionRule(Map<String, String> params) {
        acceptedViolations = params
    }

    @Override
    Violation maybeViolation(final JApiCompatibility member) {
        if (!member.binaryCompatible) {
            def acceptation = acceptedViolations[Violation.describe(member)]
            if (acceptation) {
                Violation.accept(member, acceptation)
            } else {
                Violation.notBinaryCompatible(member)
            }
        }
    }
}

and here’s how the rule is applied:

richReport {
   addRule(AcceptedRegressionRule, acceptedViolations)
}

Setup and post-process rules

Since release 0.2.2, the plugin also supports setup and post-process rules. Setup rules allow setting up some global context that can be accessed by rules extending AbstractContextAwareViolationRule. This can be useful when you need to share data between rules, and perform a final check in a post-process rule.

Setup rules need to implement SetupRule:

class MySetupRule implements SetupRule {

    @Override
    void execute(final ViolationCheckContext violationCheckContext) {
        // this is going to be executed before any other rule is executed
        violationCheckContext.userData.executed = false
    }
}

and declared using addSetupRule:

richReport {
   addSetupRule(MySetupRule)
}

Then the context can be accessed in rules implementing AbstractContextAwareViolationRule:

class ContextAwareRule extends AbstractContextAwareViolationRule {

    @Override
    Violation maybeViolation(final JApiCompatibility member) {
        // this rule is accessing the global context and can mutate user data
        context.userData.executed = true

        return null
    }
}

And then a post-process rule has access to the user data, and can also mutate the actual list of violations per class, before the report is generated:

class MyTearDownRule implements PostProcessViolationsRule {

    @Override
    void execute(final ViolationCheckContextWithViolations violationCheckContextWithViolations) {
        // this rule is executed once all checks have been performed, just before the generation
        // of the report
        // it gives the opportunity to add additional violations, or filter them, or fail
        // with a custom error
        assert violationCheckContextWithViolations.userData.executed == true
        assert !violationCheckContextWithViolations.violations.isEmpty()
    }
}

It needs to be wired in using the addPostProcessRule hook:

richReport {
   addPostProcessRule(MySetupRule)
}

wolfs/japicmp-gradle-plugin

JApicmp Gradle Plugin

Installation

Configuration

Usage

Custom reports and failure conditions

Configuration

Custom rules

Setup and post-process rules