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:
buildscript {
repositories {
jcenter()
}
dependencies {
classpath 'me.champeau.gradle:japicmp-gradle-plugin:0.2.5'
}
}
apply plugin: 'me.champeau.gradle.japicmp'
or
plugins {
id 'me.champeau.gradle.japicmp' version '0.2.5'
}
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 |
packageIncludes |
List of package names to include, * can be used as wildcard. Type: List<String> |
packageExcludes |
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 |
ignoreMissingClasses |
Ignores all superclasses or interfaces that missing on the classpath. Default value: false |
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
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")
}
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
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 |
title |
a title for the report |
description |
a description for the report |
addDefaultRules |
a boolean, indicating whether the default rules should be added or not. |
If no rules are explicitly defined, the default rules are applied. If any rule is added, the default rules won’t be applied unless addDefaultRules
is set to true
.
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
), seeJApiChangeStatus
-
on specific compatibility change descriptors (see
JApiCompatibilityChange
)
Rules are executed in the following order:
-
status change first
-
specific compatibility change
-
generic rules
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)
}
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)
}
Since 0.2.5, it is now possible to track which members have already resulted in a violation.
Since rules are executed in order, and that you can have a rule applied for a status change and a generic rule applied on the same member, it was possible for a member to trigger multiple violations.
To avoid this, you can make your rule extend AbstractRecordingSeenMembers
. This rule requires the RecordSeenMembersSetup
to be applied, and it will only add a violation, if no other violation for the same member was added before.