A Checkstyle configuration specifies which modules to plug in and apply to Java source files. Modules are structured in a tree whose root is the Checker module. The next level of modules contains:
Many checks are submodules of the TreeWalker FileSetCheck module. The TreeWalker operates by separately transforming each of the Java source files into an abstract syntax tree and then handing the result over to each of its submodules which in turn have a look at certain aspects of the tree.
Checkstyle obtains a configuration from an XML document whose elements specify the configuration's hierarchy of modules and their properties. You provide a file that contains the configuration document when you invoke Checkstyle at the command line, and when you run a Checkstyle task in ant. The documentation directory of the Checkstyle distribution contains a sample configuration file sun_checks.xml which configures Checkstyle to check for the Sun coding conventions.
A module element in the configuration XML document specifies a module identified by the element's name attribute.
Here is a fragment of a typical configuration document:
<module name="Checker">
<module name="JavadocPackage"/>
<module name="TreeWalker">
<module name="AvoidStarImport"/>
<module name="ConstantName"/>
<module name="EmptyBlock"/>
</module>
</module>
In this configuration:
For each configuration module, Checkstyle loads a class identified by the name attribute of the module. There are several rules for loading a module's class:
<module name="com.puppycrawl.tools.checkstyle.TreeWalker">
<module name="AvoidStarImport"/>
<module name="ConstantName"/>
Properties of a module control how the module performs its task. Each module property has a default value, and you are not required to define a property in the configuration document if the default value is satisfactory. To assign a non-default value to a module's property, define a child property element of the module element in the configuration XML document. Also provide appropriate name and value attributes for the property element. For example, property max of module MethodLength specifies the maximum allowable number of lines in a method or constructor, and has default value 150. Here is a configuration of module MethodLength so that the check reports methods and constructors with more than 60 lines:
<module name="MethodLength">
<property name="max" value="60"/>
</module>
Command line properties and ant Checkstyle task properties apply to the root Checker module. Also, properties are inherited in the module hierarchy. These features make it easy to define one property value that applies to many modules. For example, the following configuration fragment specifies that a tabWidth of 4 applies to TreeWalker and its submodules:
<module name="Checker">
<module name="JavadocPackage"/>
<module name="TreeWalker">
<property name="tabWidth" value="4"/>
<module name="AvoidStarImport"/>
<module name="ConstantName"/>
...
</module>
</module>
The value of a module property can be specified through property expansion with the ${property_name} notation, where property_name is a command line property or an ant Checkstyle task property. For example, the following configuration document element gives property headerFile the value of command line property checkstyle.header.file:
<property name="headerFile" value="${checkstyle.header.file}"/>
You can use property expansion to re-specify a module property value without changing the configuration document.
The property element provides an optional default attribute which is used when a property in the value cannot be resolved. For example this configuration snippet from a central configuration file checks that methods have javadoc, but allows individual projects to override the severity by specifying their desired value in the command line property checkstyle.javadoc.severity:
<module name="JavaDocMethod">
<property name="severity"
value="${checkstyle.javadoc.severity}"
default="error"/>
</module>
This feature is a great help when setting up a centralized configuration file (e.g. one file for the whole company) to lower configuration maintenance costs. Projects that are happy with the default can simply use that configuration file as is, but individual projects with special needs have the flexibility to adjust a few settings to fit their needs without having to copy and maintain the whole configuration.
All configurations have root module Checker. Checker contains:
Checker also defines properties that are inherited by all other modules of a configuration.
| name | description | type | default value |
|---|---|---|---|
| basedir | base directory name; stripped off in messages about files | string | null |
| localeCountry | locale country for messages | string: either the empty string or an uppercase ISO 3166 2-letter code | default locale country for the Java Virtual Machine |
| localeLanguage | locale language for messages | string: either the empty string or a lowercase ISO 639 code | default locale language for the Java Virtual Machine |
| charset | name of the file charset | String | System property "file.encoding" |
For example, the following configuration fragment specifies base directory src/checkstyle and German locale for all modules:
<module name="Checker">
<property name="basedir" value="src/checkstyle"/>
<property name="localeCountry" value="DE"/>
<property name="localeLanguage" value="de"/>
<module name="JavadocPackage"/>
<module name="TreeWalker">
...
</module>
</module>
To configure a Checker so that it handles files with the UTF-8 charset:
<module name="Checker">
<property name="charset" value="UTF-8"/>
...
</module>
FileSetCheck TreeWalker checks individual Java source files and defines properties that are applicable to checking such files.
| name | description | type | default value |
|---|---|---|---|
| cacheFile | caches information about files that have checked ok; used to avoid repeated checks of the same files | string | null (no cache file) |
| tabWidth | number of expanded spaces for a tab character ('\t'); used in messages and Checks that require a tab width, such as LineLength | integer | 8 |
| fileExtensions | file type extension to identify java files. Setting this property is typically only required if your java source code is preprocessed before compilation and the original files do not have the extension .java | String Set | java |
For example, the following configuration fragment specifies TreeWalker cache file target/cachefile, and a tabWidth of 4:
<module name="Checker">
<module name="TreeWalker">
<property name="cacheFile" value="target/cachefile"/>
<property name="tabWidth" value="4"/>
...
</module>
</module>
To integrate Checkstyle with BEA Weblogic Workshop 8.1:
<module name="Checker">
<module name="TreeWalker">
<property name="fileExtensions" value="java,ejb,jpf"/>
...
The TreeWalker module creates a syntax tree for a Java source file and invokes its submodules, called Checks, during a walk, or traversal, of the nodes of the tree. Every node of the syntax tree has a token. A visit to a node during the traversal triggers all Checks that are configured for its token. For example, if Check MethodLength has been configured as a submodule of TreeWalker, then a visit to a node with a method or a constructor definition token triggers MethodLength to check the number of lines of the node's code block.
Some Checks, such as FileLength and LineLength, apply directly to the source file and do not involve tokens of the syntax tree. Other Checks are associated with configurable sets of tokens that trigger the Checks. For example, this element configures Check MethodLength:
<module name="MethodLength"/>
The default token set for MethodLength is {METHOD_DEF, CTOR_DEF}, method definition and constructor definition tokens, so TreeWalker invokes MethodLength whenever it visits a node with a METHOD_DEF or a CTOR_DEF token.
You specify the trigger tokens for a Check with property tokens. The value of tokens is a list that denotes a subset of the Check's tokens, as in the following element that configures Check MethodLength to check the number of lines of methods only:
<module name="MethodLength">
<property name="tokens" value="METHOD_DEF"/>
</module>
To apply particular properties to different subsets of tokens for a Check, repeat the Check. For example, to check that the length of each method is at most 150 lines (the default value of MethodLength property max) and the length of each constructor is at most 60 lines, include the following in the TreeWalker configuration:
<module name="MethodLength">
<property name="tokens" value="METHOD_DEF"/>
</module>
<module name="MethodLength">
<property name="tokens" value="CTOR_DEF"/>
<property name="max" value="60"/>
</module>
Configurations of the Checks are specified in the pages under here.
Each check has a severity property that a Checkstyle audit assigns to all violations of the check. The default severity level of a check is error.
You can use the severity property to control the output of the plain formatter for the command line tool and the ANT task. The plain formatter does not report violations with severity level ignore, and notes violations with severity level warning. For example, according to the following configuration fragment, the plain formatter outputs warnings for translation violations:
<module name="Translation">
<property name="severity" value="warning"/>
</module>
The XML formatter reports the severity level of every violation as an attribute of the violation's error element.
As of Checkstyle 5 all checks can be configured to report custom, configuration specific messages instead of the Checkstyle default messages. This can be useful in cases where the check message should reference corresponding sections in a coding style document or the default is too generic for developers to understand.
An example usage is:
<module name="MemberName">
<property name="format" value="^m[a-zA-Z0-9]*$"/>
<message key="name.invalidPattern"
value="Member ''{0}'' must start with a lowercase ''m'' (checked pattern ''{1}'')."
/>
</module>
Each check configuration element can zero or more message elements. Every check uses one or more distinct message keys to log violations. If you want to customize a certain message you need to specify the message key in the key attribute of the message element.
The value attribute specifies the custom message pattern, as shown in the example above. Placeholders used in the default message can also be used in the custom message. Note that the message pattern must be a valid java.text.MessageFormat style pattern, so be careful about curly braces outside a placeholder definition.
The obvious question is how do you know which message keys a Check uses, so that you can override them? Well, that is the tricky part. To find out which keys a Check uses you currently need to look into the Check's source code, in conjunction with the Check's messages.properties file. Tools/plugins might come to the rescue on this topic, so have a look there.
A Checker module has a set of Filter submodules to filter audit events, including the error messages fired by Checks. A Filter can accept or reject an audit event. If all Filters accept an audit event, then the Checker reports the event. If any Filter rejects an event, then the Checker does not report the event.
Filter SeverityMatchFilter decides audit events according to the severity level of the event.
| name | description | type | default value |
|---|---|---|---|
| severity | the severity level of this filter | severity | error |
| acceptOnMatch | If acceptOnMatch is true, then the filter accepts an audit event if and only if there is a match between the event's severity level and property severity. If acceptOnMatch is false, then the filter accepts an audit event if and only if there is not a match between the event's severity level and property severity. | boolean | true |
For example, the following configuration fragment directs the Checker to not report audit events with severity level info:
<module name="SeverityMatchFilter">
<property name="severity" value="info"/>
<property name="acceptOnMatch" value="false"/>
</module>
Filter SuppressionFilter rejects audit events for Check errors according to a suppressions XML document in a file. If there is no configured suppressions file, the Filter accepts all audit events.
| name | description | type | default value |
|---|---|---|---|
| file | the location of the suppressions XML document file. The location can either be a filesystem location, or a name passed to the ClassLoader.getResource() method. | string | none |
For example, the following configuration fragment directs the Checker to use a SuppressionFilter with suppressions file docs/suppressions.xml:
<module name="SuppressionFilter">
<property name="file" value="docs/suppressions.xml"/>
</module>
A suppressions XML document contains a set of suppress elements, where each suppress element can have the following attributes:
Each audit event is checked against each suppress element. It is suppressed if all specified attributes match against the audit event.
For example, the following suppressions XML document directs a SuppressionFilter to reject JavadocStyleCheck errors for lines 82 and 108 to 122 of file AbstractComplexityCheck.java, and MagicNumberCheck errors for line 221 of file JavadocStyleCheck.java:
<?xml version="1.0"?>
<!DOCTYPE suppressions PUBLIC
"-//Puppy Crawl//DTD Suppressions 1.1//EN"
"http://www.puppycrawl.com/dtds/suppressions_1_1.dtd">
<suppressions>
<suppress checks="JavadocStyleCheck"
files="AbstractComplexityCheck.java"
lines="82,108-122"/>
<suppress checks="MagicNumberCheck"
files="JavadocStyleCheck.java"
lines="221"/>
</suppressions>
As another example, imagine that a configuration contains the following:
<module name="DescendantToken">
<property name="id" value="stringEqual"/>
<property name="tokens" value="EQUAL,NOT_EQUAL"/>
<property name="limitedTokens" value="STRING_LITERAL"/>
<property name="maximumNumber" value="0"/>
<property name="maximumDepth" value="1"/>
</module>
<module name="DescendantToken">
<property name="id" value="switchNoDefault"/>
<property name="tokens" value="LITERAL_SWITCH"/>
<property name="maximumDepth" value="2"/>
<property name="limitedTokens" value="LITERAL_DEFAULT"/>
<property name="minimumNumber" value="1"/>
</module>
Then the following can be used to suppress only the first check and not the second by using the id attribute:
<suppress id="stringEqual" files="SomeTestCode.java"/>
Filter SuppressionCommentFilter uses pairs of comments to suppress audit events.
Rationale: Sometimes there are legitimate reasons for violating a check. When this is a matter of the code in question and not personal preference, the best place to override the policy is in the code itself. Semi-structured comments can be associated with the check. This is sometimes superior to a separate suppressions file, which must be kept up-to-date as the source file is edited.
Usage: This filter only works in conjunction with a FileContentsHolder, since that check makes the suppression comments in the .java files available sub rosa. A configuration that includes this filter must configure FileContentsHolder as a child module of TreeWalker.
| name | description | type | default value |
|---|---|---|---|
| offCommentFormat | comment pattern to trigger filter to begin suppression | regular expression | CHECKSTYLE\:OFF |
| onCommentFormat | comment pattern to trigger filter to end suppression | regular expression | CHECKSTYLE\:ON |
| checkFormat | check pattern to suppress | regular expression | .* (all checks) |
| messageFormat | message pattern to suppress | regular expression | none |
| checkCPP | whether to check C++ style comments (//) | boolean | true |
| checkC | whether to check C style comments (/* ... */) | boolean | true |
offCommentFormat and onCommentFormat must have equal paren counts.
To configure the check that makes comments available to the filter:
<module name="TreeWalker">
...
<module name="FileContentsHolder"/>
...
</module>
To configure a filter to suppress audit events between a comment containing CHECKSTYLE:OFF and a comment containing CHECKSTYLE:ON:
<module name="SuppressionCommentFilter"/>
To configure a filter to suppress audit events between a comment containing line BEGIN GENERATED CODE and a comment containing line END GENERATED CODE:
<module name="SuppressionCommentFilter">
<property name="offCommentFormat" value="BEGIN GENERATED CODE"/>
<property name="onCommentFormat" value="END GENERATED CODE"/>
</module>
To configure a filter so that // stop constant check and // resume constant check marks legitimate constant names:
<module name="SuppressionCommentFilter">
<property name="offCommentFormat" value="stop constant check"/>
<property name="onCommentFormat" value="resume constant check"/>
<property name="checkFormat" value="ConstantNameCheck"/>
</module>
To configure a filter so that UNUSED OFF: var and UNUSED ON: var marks a variable or parameter known not to be used by the code by matching the variable name in the message:
<module name="SuppressionCommentFilter">
<property name="offCommentFormat" value="UNUSED OFF\: (\w+)"/>
<property name="onCommentFormat" value="UNUSED ON\: (\w+)"/>
<property name="checkFormat" value="Unused"/>
<property name="messageFormat" value="^Unused \w+ '$1'.$"/>
</module>
To configure a filter so that CSOFF: regexp and CSN: regexp mark a matching check:
<module name="SuppressionCommentFilter">
<property name="offCommentFormat" value="CSOFF\: ([\w\|]+)"/>
<property name="onCommentFormat" value="CSON\: ([\w\|]+)"/>
<property name="checkFormat" value="$1"/>
</module>
To configure a filter to suppress all audit events between a comment containing CHECKSTYLE_OFF: ALL and a comment containing CHECKSTYLE_OFF: ALL except for the EqualsHashCode check:
<module name="SuppressionCommentFilter">
<property name="offCommentFormat" value="CHECKSTYLE_OFF: ALL"/>
<property name="onCommentFormat" value="CHECKSTYLE_ON: ALL"/>
<property name="checkFormat" value="^((?!(EqualsHashCode)).)*$"/>
</module>
Filter SuppressWithNearbyCommentFilter uses individual comments to suppress audit events.
Rationale: Same as SuppressionCommentFilter. Whereas the SuppressionCommentFilter uses matched pairs of filters to turn on/off comment matching, SuppressWithNearbyCommentFilter uses single comments. This requires fewer lines to mark a region, and may be aesthetically preferable in some contexts.
Usage: This filter only works in conjunction with a FileContentsHolder, since that check makes the suppression comments in the .java files available sub rosa. A configuration that includes this filter must configure FileContentsHolder as a child module of TreeWalker.
| name | description | type | default value |
|---|---|---|---|
| commentFormat | comment pattern to trigger filter to begin suppression | regular expression | SUPPRESS CHECKSTYLE (\w+) |
| checkFormat | check pattern to suppress | regular expression | .* |
| messageFormat | message pattern to suppress | regular expression | none |
| influenceFormat | a negative/zero/positive value that defines the number of lines preceding/at/following the suppression comment | regular expression | 0 (the line containing the comment) |
| checkCPP | whether to check C++ style comments (//) | boolean | true |
| checkC | whether to check C style comments (/* ... */) | boolean | true |
To configure the check that makes comments available to the filter:
<module name="TreeWalker">
...
<module name="FileContentsHolder"/>
...
</module>
To configure a filter to suppress audit events for check on any line with a comment SUPPRESS CHECKSTYLE check:
<module name="SuppressWithNearbyCommentFilter"/>
To configure a filter to suppress all audit events on any line containing the comment CHECKSTYLE IGNORE THIS LINE:
<module name="SuppressWithNearbyCommentFilter">
<property name="commentFormat" value="CHECKSTYLE IGNORE THIS LINE"/>
<property name="checkFormat" value=".*"/>
<property name="influenceFormat" value="0"/>
</module>
To configure a filter so that // Ok to catch (Throwable|Exception|RuntimeException) here permits the current and previous line to avoid generating an IllegalCatch audit event:
<module name="SuppressWithNearbyCommentFilter">
<property name="commentFormat" value="Ok to catch (\w+) here"/>
<property name="checkFormat" value="IllegalCatchCheck"/>
<property name="messageFormat" value="$1"/>
<property name="influenceFormat" value="-1"/>
</module>
To configure a filter so that CHECKSTYLE IGNORE check FOR NEXT var LINES avoids triggering any audits for the given check for the current line and the next var lines (for a total of var+1 lines):
<module name="SuppressWithNearbyCommentFilter">
<property name="commentFormat" value="CHECKSTYLE IGNORE (\w+) FOR NEXT (\d+) LINES"/>
<property name="checkFormat" value="$1"/>
<property name="influenceFormat" value="$2"/>
</module>
In addition to an audit reporter for text or XML output, a Checker can have custom AuditListeners that handle audit events. In order to use a custom listener, add a Checker submodule for the listener and its properties. For example, to configure a Checker so that it uses custom listener VerboseListener to print audit messages to a file named "audit.txt", include the following module in the configuration file:
<module name="com.mycompany.listeners.VerboseListener">
<property name="file" value="audit.txt"/>
</module>
Checkstyle loads a module class according to the name of a module element, and automatically appends pre-specified package prefixes to that name in its search for a loadable class. By default, Checkstyle applies packages com.puppycrawl.tools.checkstyle, com.puppycrawl.tools.checkstyle.filters, and com.puppycrawl.tools.checkstyle.checks as well as any sub-packages of com.puppycrawl.tools.checkstyle.checks that are distributed with Checkstyle.
To specify other packages to apply, create a package names XML document in a file named checkstyle_packages.xml, and provide that file in the root of the .jar containing your custom checks.
Note that the support for providing a package names XML document via command line option or as a attribute of an ant Checkstyle task has been dropped with Checkstyle 5.0.
A package names XML document specifies a list of package names. Here is a sample package names XML document for packages com.puppycrawl.tools.checkstyle and com.puppycrawl.tools.checkstyle.checks:
<checkstyle-packages>
<package name="com.puppycrawl.tools.checkstyle">
<package name="checks"/>
</package>
</checkstyle-packages>
Notice that the packages are specified recursively - a child package element is a subpackage of its parent package element.
For example, to incorporate modules from package com.mycompany.checks with Checkstyle modules, create the XML file below and put this file into the root of the jar file which contains your custom check modules. The XML file must be named exactly checkstyle_packages.xml:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE checkstyle-packages PUBLIC
"-//Puppy Crawl//DTD Package Names 1.0//EN"
"http://www.puppycrawl.com/dtds/packages_1_0.dtd">
<checkstyle-packages>
<package name="com.mycompany.checks"/>
</checkstyle-packages>
Now you can configure a module of package com.mycompany.checks, say com.mycompany.checks.MethodLimitCheck, with a shortened module element in the configuration document:
<module name="MethodLimit"/>
The following DTD for a configuration XML document specifies the hierarchy of modules and their properties:
<?xml version="1.0" encoding="UTF-8"?>
<!ELEMENT module (module|property)*>
<!ATTLIST module name NMTOKEN #REQUIRED>
<!ELEMENT property EMPTY>
<!ATTLIST property
name NMTOKEN #REQUIRED
value CDATA #REQUIRED
>
Checkstyle validates a configuration XML document when it loads the document. To validate against the above configuration DTD, include the following document type declaration in your configuration XML document:
<!DOCTYPE module PUBLIC
"-//Puppy Crawl//DTD Check Configuration 1.3//EN"
"http://www.puppycrawl.com/dtds/configuration_1_3.dtd">
Checkstyle also strictly enforces the encoding attribute of an XML declaration. If Checkstyle rejects your configuration document's encoding, correct the value of the encoding attribute, or remove the encoding attribute entirely.
For a complete example of a configuration XML document, examine file docs/sun_checks.xml that checks the Sun coding conventions.
This is a DTD for a package names XML document:
<?xml version="1.0" encoding="UTF-8"?>
<!ELEMENT checkstyle-packages (package)*>
<!ELEMENT package (package)*>
<!ATTLIST package name NMTOKEN #REQUIRED>
Checkstyle also validates a package names XML document when it loads the document. To validate against the above package names DTD, include the following document type declaration in your package names XML document:
<!DOCTYPE checkstyle-packages PUBLIC
"-//Puppy Crawl//DTD Package Names 1.1//EN"
"http://www.puppycrawl.com/dtds/packages_1_1.dtd">