Java
On this page
Language-specific properties
You can discover and update the Java-specific properties in Administration > General Settings > Languages > Java.
Java analysis and bytecode
Compiled .class
files are required for java projects with more than one java file. If not provided properly, analysis will fail with the message:
If only some .class
files are missing, you'll see warnings like this:
If you are not using Maven or Gradle for analysis, you must manually provide bytecode to the analysis. You can also analyze test code, and for that you need to provide tests binaries and test libraries properties.
Key | Value |
sonar.java.binaries (required) | Comma-separated paths to directories containing the compiled bytecode files corresponding to your source files. |
sonar.java.libraries | Comma-separated paths to files with third-party libraries (JAR or Zip files) used by your project. Wildcards can be used: sonar.java.libraries=path/to/Library.jar,directory/**/*.jar |
sonar.java.test.binaries | Comma-separated paths to directories containing the compiled bytecode files corresponding to your test files |
sonar.java.test.libraries | Comma-separated paths to files with third-party libraries (JAR or Zip files) used by your tests. (For example, this should include the junit jar). Wildcards can be used: sonar.java.test.libraries=directory/**/*.jar |
Android users, Jack doesn't provide the required .class
files.
Project's specific JDK
In some situations, you might have to analyze a project built with a different version of Java than the one executing the analysis. The most common case is to run the analysis with Java 17, while the project itself uses Java 11 or before for its build.
If this is your case, you will need to set the sonar.java.jdkHome
property manually to point the appropriate JDK (see below). By doing this you will specify which JDK classes the analyzer must refer to during the analysis. Not setting this property, while it would have been required, usually leads to inconsistent or even impossible to fix issues being reported, especially in relation with native JDK classes.
When setting sonar.java.jdkHome
, you need to provide the path to the JDK directory used by the project being analyzed, if different from the Java runtime executing the analysis. For example, for a Java 11 project, by setting it as follows: sonar.java.jdkHome=/usr/lib/jvm/jdk11/
This option can of course be added to your sonar.properties
configuration.
JDK preview features
To enable the JDK preview features in SonarQube, you can set the sonar.java.enablePreview
analysis parameter to true
(default is false
).
Turning issues off
The best way to deactivate an individual issue you don't intend to fix is to mark it "Won't Fix" or "False Positive" through the SonarQube UI.
If you need to deactivate a rule (or all rules) for an entire file, then issue exclusions are the way to go. But if you only want to deactivate a rule across a subset of a file - all the lines of a method or a class - you can use @SuppressWarnings("all")
or @SuppressWarnings
with rule keys: @SuppressWarnings("java:S2077")
or @SuppressWarnings({"java:S1118", "java:S3546"})
.
Handling Java source version
Java analysis is able to react to the java version used for sources. This feature allows the deactivation of rules that target higher versions of Java than the one in use in the project so that false positives aren't generated from irrelevant rules.
The feature relies entirely on the sonar.java.source
property, which is automatically filled by most of the scanners used for analyses (Maven, Gradle). Java version-specific rules are not disabled when sonar.java.source
is not provided. Concretely, rules which are designed to target specific java versions (tagged "java7" or "java8") are activated by default in the Sonar Way Java profile. From a user perspective, the feature is fully automatic, but it means that you probably want your projects to be correctly configured.
When using SonarScanner to perform analyses of project, the property sonar.java.source
can to be set manually in sonar-project.properties
. Accepted formats are:
- "1.X" (for instance 1.6 for java 6, 1.7 for java 7, 1.8 for java 8, etc.)
- "X" (for instance 7 for java 7, 8 for java 8, etc. )
Example: sonar.java.source=1.6
If the property is provided, the analysis will take the source version into account, and execute related rules accordingly. At run time, each of these rules will be executed – or not – depending of the Java version used by sources within the project. For instance, on a correctly configured project built with Java 6, rules targeting Java 7 and Java 8 will never raise issues, even though they are enabled in the associated rule profile.
Batch mode settings
By default, files are parsed in batches. The size of the batch is dynamically computed based on the maximum memory available. It is possible to manually set this value by using the property sonar.java.experimental.batchModeSizeInKB
. Note that the perfect value depends on the project and the ecosystem setup, bigger batch size will not necessarily increase the performance and can even slow things down if the memory is a limiting factor. If needed, it is possible to run the parsing file by file by setting sonar.java.fileByFile=true
.
More details can be found here.
Skipping unchanged files
Starting from April 2022, and by default, the Java analyzer optimizes the analysis of unchanged files in pull requests. In practice, this means that on a file that has not changed, the analyzer only runs a restricted list of rules. To get a better understanding of the rule exclusion mechanism, keep in mind that:
- Rules that need to run on multiple files to decide whether they need to raise issues are always executed
- Rules that need to run at the end of the analysis to decide whether they need to raise issues are always executed
- Rules that are defined outside of the
sonar.java.checks
package are always executed
This last criteria implies that custom rules cannot be skipped.
If you wish to disable this optimization, you can set the value of the analysis parameter sonar.java.skipUnchanged
to false
. Leaving the parameter unset lets the server decide whether the optimization should be enabled.
Cache-enabled rules (experimental)
Starting from April 2022, the Java analyzer offers rule developers a SQ cache that can be used to store and retrieve information from one analysis to the other. The cache is provided by the underlying SonarQube instance and is branch-specific. Please refer to the sonar-java wiki for additional information.
Analyzing JSP and Thymeleaf for XSS vulnerabilities
In SonarQube Developer and Enterprise editions and on SonarCloud you can benefit from advanced security rules including XSS vulnerability detection. Java analysis supports analysis of Thymeleaf and JSP views when used with Java Servlets or Spring. To benefit from this analysis you need to make your views part of the project sources using sonar.sources
property. In practice this usually means adding the following in your Maven pom.xml
file
or if you use Gradle
where src/main/webapp
is the directory which contains .jsp
or Thymeleaf's .html
files.
Implementation Related Rule Tags
- symbolic-execution: This tag is for rules that reason about the state of the program using data flow analysis. They usually work together to find path-sensitive bugs and vulnerabilities. As soon as an issue is raised, the symbolic execution (SE) analysis of the current path will stop. For that reason, it is not recommended to evaluate these rules independently of each other as it can give a false sense of undetected issues. It is important to keep in mind that SE can never achieve perfection, so we are always working on improving these rules. Finally, note that the Java rules relying on the SE engine operate cross-procedurally in certain circumstances. In particular, all non-overridable methods defined in the same file as the method under analysis and called from within the method's body, will be explored and learned from. Behaviors of overridable methods will be approximated.
Related pages
- Test coverage and execution (JaCoCo, Surefire)
- Importing external issues (SpotBugs, FindBugs, FindSecBugs, PMD, Checkstyle)
- Adding coding rules
Custom rules
The tutorial Writing custom Java rules 101 will help to quickly start writing custom rules for Java.
API changes
7.25
Update custom rules registration API CheckRegistrar.RegistrarContext
to allow:
- Registration of JavaCheck instances (previously only classes were supported)
registerMainChecks(String repositoryKey, Collection<?> javaCheckClassesAndInstances)
registerTestChecks(String repositoryKey, Collection<?> javaCheckClassesAndInstances)
- Registration of a JavaCheck that has no
@Rule
annotation because it does not target one but several rules.registerMainSharedCheck(JavaCheck check, Collection<RuleKey> ruleKeys)
registerTestSharedCheck(JavaCheck check, Collection<RuleKey> ruleKeys)
- Deprecate useless getters from the RegistrarContext class and add
TestCheckRegistrarContext
for test purposes.
Add custom rules registration API ProfileRegistrar.RegistrarContext
to allow:
- Registration of rules in the "Sonar Way" built-in default quality profile for the Java language.
registerDefaultQualityProfileRules(Collection<RuleKey> ruleKeys)
7.19
All the API changes are related to the support of the preview feature of Java 19/20. These new types and methods are introduced as "deprecated" and will be marked out of deprecation with the introduction of the final features in the JDK.
- New type:
RecordPatternTree
. Use this type to explore record patterns. - New method:
RecordPatternTree#type()
. Use this method to get the reference type in the record pattern. - New method:
RecordPatternTree#patterns()
. Use this method to get the patterns nested in the record pattern. - New method:
RecordPatternTree#name()
. Use this method get the optional record pattern identifier. - Dropped method:
GuardedPatternTree#andOperator()
. UseGuardedPatternTree#whenOperator()
instead. The&&
operator from Java 18 got replaced by a new restricted keywordwhen
in Java 19. - New method:
GuardedPatternTree#whenOperator()
. Use this method to get thewhen
operator prefixing the guard. - Deprecated method:
PatternInstanceOfTree#variable()
. UsePatternInstanceOfTree#pattern()
instead. Java 19 introduced record patterns and their support in the newinstanceof
. - New method:
PatternInstanceOfTree#pattern()
. Use this method to get the pattern in aninstanceof
expression. When the pattern is aTypePatternTree
, the variable can then be extracted usingTypePatternTree#patternVariable()
. - New method:
TreeVisitor#visitRecordPattern()
. Use this method to traverse aRecordPatternTree
.
7.17
- New method:
ClassTree#recordOpenParenToken()
. Use this method to get the opening parenthesis from the component list of arecord
declaration. - New method:
ClassTree#recordCloseParenToken()
. Use this method to get the closing parenthesis from the component list of arecord
declaration.
7.16
- New method:
Position#startOf(Tree)
. Use this static method to get the start position of a givenTree
. Shortcuts the usually cumbersome operation of accessing a tree start position through the normal API, withtree.firstToken().range().start()
. - New method:
Position#startOf(SyntaxToken)
. Use this static method to get the start position of a givenSyntaxToken
. - New method:
Position#startOf(SyntaxTrivia)
. Use this static method to get the start position of a givenSyntaxTrivia
. - New method:
Position#endOf(Tree)
. Use this static method to get the end position of a givenTree
. Shortcuts the usually cumbersome operation of accessing a tree end position through the normal API, withtree.lastToken().range().end()
. - New method:
Position#endOf(SyntaxToken)
. Use this static method to get the end position of a givenSyntaxToken
. - New method:
Position#endOf(SyntaxTrivia)
. Use this static method to get the end position of a givenSyntaxTrivia
. - New method:
MethodInvocationTree#methodSymbol()
. Use this method to get the symbol of a method invocation. Returns aMethodSymbol
. - New method:
NewClassTree#methodSymbol()
. Use this method to get the symbol of a constructor call. Returns aMethodSymbol
. - Deprecated method:
MethodInvocationTree#symbol()
. Deprecated in favor ofmethodSymbol()
, which has a narrower return type. - Deprecated method:
NewClassTree#constructorSymbo()
. Deprecated in favor ofmethodSymbol()
, which has a narrower return type.
7.15
- New method:
JavaResourceLocator#binaryDirs()
. Use this method to get the directories containing the .class files corresponding to the main code. - New method:
JavaResourceLocator#testBinaryDirs()
. Use this method to get the directories containing the .class files corresponding to the tests. - New method:
JavaResourceLocator#testClasspath()
. Use this method to retrieve the classpath configured for the project's tests.
7.12
- New method:
JavaFileScanner#scanWithoutParsing(InputFileScannerContext)
. Use this method to inspect an unchanged file before it is parsed. This method allows you to pre-compute some work and even signal that the rule does not need the file to be parsed. TypeArguments
extendsListTree<TypeTree>
instead ofListTree<Tree>
. This change does not impact the runtime compatibility but could break compile time compatibility in some rare cases.
7.7
- Method
MethodSymbol.overriddenSymbol()
was dropped.. Deprecated in 6.15 and planned to be dropped in 7.0,overriddenSymbol()
has now been removed from the API. UseMethodSymbol.overriddenSymbols()
instead.
7.6
- New method:
MethodSymbol.declarationParameters()
. Use this method to get the list of parameters symbols of this method. Placeholders symbols are created in case the declaration is not available (coming from an external dependency).
7.4
- New method:
JavaFileScannerContext.inAndroidContext()
. Use this method to know if the current file being analyzed is coming from an Android context. The value is true if Android dependencies are found in the classpath of the current analysis. - Together with the previous method addition, you can use
CheckVerifier.withinAndroidContext(true)
in unit tests to test the behavior of the rules in an Android context.
7.1
- Method
TypeCastTree.bounds()
changed its return type fromListTree<Tree>
toListTree<TypeTree>
. Having any kind of tree is not possible. Only "Typed" Trees are possible as bound of a cast. The change fixes the inconsistency. - Method
TypeParameterTree.bounds()
changed its return type fromListTree<Tree>
toListTree<TypeTree>
. Having any kind of tree is not possible. Only "Typed" Trees are possible as bound of a type parameter. The change fixes the inconsistency. - A new Tree Kind
PATTERN_INSTANCE_OF
has been formalized, with its new corresponding API classPatternInstanceOfTree
. The change follows Java 16 introduction of Pattern Matching forinstanceof
(JEP-394). - A new Tree Kind
RECORD
has been formalized, adding a flavor toClassTree
. The change follows Java 16 introduction ofrecord
s (JEP-395).
7.0
- Method
MethodTree.isOverriding()
now correctly match the contract in case of unknowns in the hierarchy. Previously, theisOverriding
implementation could return misleading results when the hierarchy of the class that the method belonged to contained unknowns. In such cases, null is now returned, because it is not possible to reliably determine if the method is an override or not.
The following deprecated methods have been dropped:
org.sonar.plugins.java.api.JavaFileScannerContext.getFileKey()
The method was deprecated for a long time and still used by Sonar-Security. It's not the case anymore, and the method can be dropped, without replacement.org.sonar.plugins.java.api.SourceMap.Location.inputFile()
The method has been deprecated a few versions away and was only used by Sonar-Security. It's not used anymore, and the method can be dropped, without replacement.org.sonar.plugins.java.api.semantic.Symbol.MethodSymbol.overriddenSymbol()
The method has been replaced byMethodSymbol.overriddenSymbols()
, which provide a list of overridden methods instead of only one.org.sonar.plugins.java.api.tree.BreakStatementTree.value()
The method was deprecated in SonarJava 6.6. It was added to cover new switch expression of Java 12. The new switch expression was introduced officially in Java 14, using a newyield
statement instead of relying onbreak
statement, leading to the removal of this method.org.sonar.plugins.java.api.tree.CaseLabelTree.expression()
The method was deprecated since SonarJava 5.12 when introducing support of Java 12, in favor ofCaseLabelTree.expressions()
method.org.sonar.plugins.java.api.tree.CaseLabelTree.colonToken()
The method was deprecated since SonarJava 5.12 when introducing support of Java 12, in favor ofCaseLabelTree.colonOrArrowToken()
method.org.sonar.plugins.java.api.tree.SwitchStatementTree.asSwitchExpression()
The method was making no sense and starting from SonarJava 6.15, switch expressions have their own distinct interface (org.sonar.plugins.java.api.tree.SwitchExpressionTree
) and kind (org.sonar.plugins.java.api.tree.Tree.Kind.SWITCH_EXPRESSION
).
The following classes have been dropped, since all their methods where already deprecated:
- Classes
org.sonar.java.checks.verifier.JavaCheckVerifier
andorg.sonar.java.checks.verifier.MultipleFilesJavaCheckVerifier
. All features are new achievable throughCheckVerifier.newVerifier()
.
6.15
- Switch representation change in the AST Previously, Switch Statements were represented thanks to a Switch Expression. It means that the child of the Switch Statement was a Switch Expression. This is no longer the case, a Switch statement is now a distinct node in the AST and does not contain a Switch Expression anymore. This may impact existing custom rules relying explicitly on Switch Expressions, via the kind SWITCH_EXPRESSION or the method visitSwitchExpression. More rarely, this could also impact rules relying on parents of Tree, as the overall shape of the AST may also change.
- Deprecated
org.sonar.plugins.java.api.tree.SwitchStatementTree
: TheasSwitchExpression()
method is deprecated for removal.org.sonar.plugins.java.api.semantic.Symbol.MethodSymbol
: TheoverriddenSymbol()
method is deprecated for removal. It is replaced by methodoverriddenSymbols()
which returns all the overridden symbols in the type hierarchy instead of only the first one found.
- New interface
SwitchTree
Switch Expression and Switch Statement share the same fields, it can sometimes make sense to manipulate a Switch as either one. Previously, it was possible to use the method asSwitchExpression() to do this. Since this method is deprecated, you can now use the new common interfaceSwitchTree
, containing all the elements shared between Switch Expressions and Statements.
6.3
- API is now enriched with
MethodMatchers
. You can use it to identify a method with given a Type, Name and Parameters. We realized that MethodMatchers is a really convenient way of writing new rules, it will hopefully ease the addition of rules in custom plugins, without having to rewrite the logic. We are heavily using it in the different checks, plenty of examples can be found there. - Two new methods have been added in the semantic API in order to access parametrized type in custom rules. The changes are available in
org.sonar.plugins.java.api.semantic.Type
:
- The
JavaCheckVerifier
, used to test rules implementations and delivered with thejava-checks-testkit
package, has been fully reworked in order to tackle inconcistencies. All the previously existing methods from it has been deprecated. In addition, a new method has been added, which allows access to a new rule testing interface. Starting from 6.3, when writting custom rules test, you should therefore rely only onorg.sonar.java.checks.verifier.JavaCheckVerifier.newVerifier()
. Example of change:
6.1
- The
ExpressionTree
interface, from the AST API, is now enriched by two new methodsOptional<Object> asConstant()
and<T> Optional<T> asConstant(Class<T> type)
. These methods let you try to retrieve the equivalent constant value of an expression (from a variable, for instance). An example of usage would be:
6.0
- Deprecated method
org.sonar.plugins.java.api.JavaFileScannerContext.addIssue(File, JavaCheck, int, String)
has been removed. Custom rules relying on it should report issues on a givenTree
from now on. - Deprecated method
org.sonar.plugins.java.api.JavaFileScannerContext.getFile()
has been removed. Custom rules relying on it should rely on content of SQ's APIInputFile
. - Deprecated method
org.sonar.plugins.java.api.tree.TryStatementTree.resources()
has been removed, in favor oforg.sonar.plugins.java.api.tree.TryStatementTree.resourceList()
, as Java 9 allows other trees thanVariableTree
to be placed as resources in try-with-resources statements. - Method
org.sonar.plugins.java.api.semantic.Symbol.owner()
has been flagged with@Nullable
annotation, to explicitly document the fact that some symbols (package, unknown, recovered) might well returnnull
. - Semantic engine
- Return type of constructor is now
void type
instead ofnull
. - A raw type is now explicitly different from an erasure type. It is recommended to systematically use type erasure for type comparison when dealing with generics.
- Return type of constructor is now
- According to Java Language Specification every array type implements the interface
java.io.Serializable
, callingisSubtypeOf("java.io.Serializable")
on an array type now consistently returnstrue
. - Symbol corresponding to generic method invocations are now correctly parameterized.
- In some special cases (mostly missing bytecode dependencies, misconfigured projects), and due to ECJ recovery system, unknown/recovered types can now lead to unknown symbols, even on
ClassTree
/MethodTree
/VariableTree
. To illustrate this, the following example now associate the method to an unknown symbol, while previous semantic engine from version 5.X series was creating aSymbol.MethodSymbol
with an unknown return type.
- Thanks to improved semantic provided by ECJ engine, new semantic is now able to say that an unknown symbol is supposed to be type/variable/method (
isTypeSymbol()
,isVariableSymbol()
, ...). Old semantic was answeringfalse
for all of them. Consequently, be sure to always useisUnknown()
to validate symbol resolution. Otheris...Symbol()
methods are only designed to know how to cast the symbols (e.g fromSymbol
toSymbol.MethodSymbol
).
5.12
- Dropped
org.sonar.plugins.java.api.JavaFileScannerContext
: Drop deprecated method used to retrieve trees contributing to the complexity of a method from (deprecated since SonarJava 4.1).
org.sonar.plugins.java.api.JavaResourceLocator
: The following method has been dropped (deprecated since SonarJava 4.1), without replacement.
org.sonar.plugins.surefire.api.SurefireUtils
: Dropping deprecated field with old property (deprecated since SonarJava 4.11)
- Deprecated
org.sonar.plugins.java.api.JavaFileScannerContext
: Deprecate usage of File-based methods from API, which will be removed in future release. Starting from this version, methods relying on InputFile has to be preferred.
- Deprecate methods which are not relevant anymore in switch-related trees from API, following introduction of the new Java 12
switch
expression:
- Added
org.sonar.plugins.java.api.JavaFileScannerContext
: Following methods have been added in order to provide help reporting issues at project level, and access data through SonarQube's InputFile API, which won't be possible anymore through files:
- In order to cover the Java 12 new switch expression, introduce a new Tree in the SonarJava Syntax Tree API (Corresponding
Tree.Kind
:SWITCH_EXPRESSION
). New methods have also been added to fluently integrate the new switch expression into the SonarJava API.
5.7
- Breaking
- This change will impact mostly the custom rules relying on semantic API. The type returned by some symbols will change from raw type to parameterized type with identity substitution and this will change how subtyping will answer. It is possible to get the previous behavior back by using type erasure on the newly returned type. Note that not all returned types are impacted by this change. Example:
Was this page helpful?