简介:Static code analysis for Kotlin
更多:作者   提 Bug   

build status build status windows Download All Contributors

All Contributors Awesome Kotlin Badge

Meet detekt, a static code analysis tool for the Kotlin programming language. It operates on the abstract syntax tree provided by the Kotlin compiler.

detekt in action


  • code smell analysis for your kotlin projects
  • complexity report based on logical lines of code, McCabe complexity and amount of code smells
  • highly configurable (rule set or rule level)
  • suppress findings with Kotlin's @Suppress and Java's @SuppressWarnings annotations
  • specify code smell thresholds to break your build or print a warning
  • format your code with the formatting rule set
  • code Smell baseline and ignore lists for legacy projects
  • gradle plugin for code analysis, formatting and import migration
  • gradle tasks to use local intellij distribution for formatting and inspecting kotlin code
  • optionally configure detekt for each sub module by using profiles (gradle-plugin)
  • sonarqube integration
  • NEW extensible by own rule sets and FileProcessListener's

Table of contents

  1. Commandline interface
  2. Gradle plugin
    1. in groovy dsl
    2. in kotlin dsl
    3. plugin tasks
    4. detekt-closure
  3. Standalone gradle task
  4. Standalone maven task
  5. Rule sets
  6. Rule set configuration
  7. Suppress rules
  8. Build failure
  9. Extending detekt
    1. RuleSets
    2. Processors
    3. Reports
    4. Rule testing
  10. Formatting - Code Style
  11. Black- and Whitelist code smells
  12. Contributors

Build & use the commandline interface

  • git clone
  • cd detekt
  • ./gradlew build shadow
  • java -jar detekt-cli/build/libs/detekt-cli-[version]-all.jar [parameters]*
Parameters for CLI

The CLI uses jcommander for argument parsing. Following parameters are shown when --help is entered:

The following option is required: --input, -i

Usage: detekt [options]
    --baseline, -b
      Treats current analysis findings as a smell baseline for further detekt
      runs. If a baseline xml file is passed in, only new code smells not in
      the baseline are printed in the console.
    --config, -c
      Path to the config file (path/to/config.yml).
    --config-resource, -cr
      Path to the config resource on detekt's classpath (path/to/config.yml).
    --create-baseline, -cb
      Treats current analysis findings as a smell baseline for further detekt runs.
      Default: false
    --debug, -d
      Debugs given ktFile by printing its elements.
      Default: false
    --disable-default-rulesets, -dd
      Disables default rule sets.
      Default: false
    --filters, -f
      Path filters defined through regex with separator ';' (".*test.*").
      Enables formatting of source code. Cannot be used together with
      Default: false
    --generate-config, -gc
      Export default config to default-detekt-config.yml.
      Default: false
    --help, -h
      Shows the usage.
    --output, -o
      Specify the file to output to.
    --output-name, -on
      The base name for output reports is derived from this parameter.
      Enables parallel compilation of source files. Should only be used if the
      analyzing project has more than ~200 kotlin files.
      Default: false
  * --input, -i
      Input path to analyze (path/to/project).
    --rules, -r
      Extra paths to ruleset jars separated by ';'.
      Tells the formatter that indentation with tabs are valid.
      Default: false

--input can either be a directory or a single Kotlin file. The currently only supported configuration format is yaml. --config should point to one. Generating a default configuration file is as easy as using the --generate-config parameter. filters can be used for example to exclude all test directories. With rules you can point to additional ruleset.jar's creating by yourself or others. More on this topic see section Custom RuleSets.

Using the detekt-gradle-plugin

Use the groovy or kotlin dsl of gradle and configure the detekt closure as described here.

Configuration when using groovy dsl

For gradle version >= 2.1

buildscript {
    repositories {

plugins {
    id "io.gitlab.arturbosch.detekt" version "1.0.0.[version]"

detekt {
    version = "1.0.0.[version]"
    profile("main") {
        input = "$projectDir/src/main/kotlin"
        config = "$projectDir/detekt.yml"
        filters = ".*test.*,.*/resources/.*,.*/tmp/.*"

For all gradle versions:

buildscript {
  repositories {
    maven { url "" }
  dependencies {
    classpath "[version]"

apply plugin: "io.gitlab.arturbosch.detekt"

detekt {
    version = "1.0.0.[version]"
    profile("main") {
        input = "$projectDir/src/main/kotlin"
        config = "$projectDir/detekt.yml"
        filters = ".*test.*,.*/resources/.*,.*/tmp/.*"
Configuration when using kotlin dsl

For gradle version >= 3.5

import io.gitlab.arturbosch.detekt.DetektExtension

buildscript {
    repositories {
plugins {

configure<DetektExtension> {
    version = "1.0.0.[version]"
    profile("main") {
        input = "$projectDir/src/main/kotlin"
        config = "$projectDir/detekt.yml"
        filters = ".*test.*,.*/resources/.*,.*/tmp/.*"

Available plugin tasks

  • detektCheck - Runs a detekt analysis and complexity report. Configure the analysis inside the detekt-closure. By default the standard rule set is used without output report or black- and whitelist checks.
  • detektGenerateConfig - Generates a default detekt config file into your projects location.
  • detektFormat - Formats your kotlin code by using the formatting rule set. Specify which rules should be turned on/off and which should be auto corrected. (see Formatting - Code Style). Pay attention that this formatting rule set is not as powerful as intellij's inspection and never will be. It is recommend to use the detektIdeaFormat task which needs some pre configurations.
  • detektBaseline - Like detektCheck, but creates a code smell baseline. Further detekt runs will only feature new smells not in this list.
  • detektMigrate - Experimental feature for now. Just supports migration of specified imports. See migration section.
  • detektIdeaFormat - Uses a local idea installation to format your kotlin (and other) code according to the specified code-style.xml.
  • detektIdeaInspect Uses a local idea installation to run inspections on your kotlin (and other) code according to the specified inspections.xml profile.
Options for detekt configuration closure
detekt {
    version = "1.0.0.[version]"  // When unspecified the latest detekt version found, will be used. Override to stay on the same version.

     // A profile basically abstracts over the argument vector passed to detekt. 
     // Different profiles can be specified and used for different sub modules or testing code.
    profile("main") {
        input = "$projectDir/src/main/kotlin" // Which part of your project should be analyzed?
        config = "$projectDir/detekt.yml" // Use $project.projectDir or to navigate inside your project 
        configResource = "/detekt.yml" // Use this parameter instead of config if your detekt yaml file is inside your resources. Is needed for multi project maven tasks.
        filters = ".*test.*, .*/resources/.*" // What paths to exclude? Use comma oder semicolon to separate
        ruleSets = "other/optional/ruleset.jar" // Custom rule sets can be linked to this, use comma oder semicolon to separate, remove if unused.
        disableDefaultRuleSets = false // Disables the default rule set. Just use detekt as the detection engine with your custom rule sets.
        output = "$project.projectDir/reports/detekt.xml" // If present, prints all findings into that file.
        outputName = "my-module" // This parameter is used to derive the output report name
        baseline = "$project.projectDir/reports/baseline.xml" // If present all current findings are saved in a baseline.xml to only consider new code smells for further runs.
        parallel = true // Use this flag if your project has more than 200 files. 
        useTabs = false // Turns off the indentation check for spaces if true, default is false and does not need to be specified

   // Definines a secondary profile `gradle detektCheck -Ddetekt.profile=override` will use this profile. 
   // The main profile gets always loaded but specified profiles override main profiles parameters.
   profile("override") {
       config = "$projectDir/detekt-test-config.yml"
Configure a local idea for detekt
  • download the community edition of intellij
  • extract the file to your preferred location eg. ~/.idea
  • let detekt know about idea inside the detekt-closure
  • extract code-style.xml and inpect.xml from idea settings (Settings>CodeStyle>Scheme and Settings>Inspections>Profile)
  • run detektIdeaFormat or detektIdeaInspect
  • all parameters in the following detekt-closure are mandatory for both tasks
  • make sure that current or default profile have an input path specified!
String USER_HOME = System.getProperty("user.home")

detekt {  
    profile("main") {
        input = "$projectDir/src/main/kotlin"
        output = "$projectDir/reports/report.xml"
        outputFormat = "xml"
    idea {
        path = "$USER_HOME/.idea"
        codeStyleScheme = "$USER_HOME/.idea/idea-code-style.xml"
        inspectionsProfile = "$USER_HOME/.idea/inspect.xml"
        report = "project.projectDir/reports"
        mask = "*.kt,"

For more information on using idea as a headless formatting/inspection tool see here.


Migration rules can be configured in your detekt.yml file. For now only migration of imports is supported.

# *experimental feature*
# Migration rules can be defined in the same config file or a new one
  active: true
    # your.package.Class: new.package.or.Class
    # for example:
    # io.gitlab.arturbosch.detekt.api.Rule: io.gitlab.arturbosch.detekt.rule.Rule

Using detekt in custom gradle projects

  1. Add following lines to your build.gradle file.
  2. Run gradle detekt
  3. Add check.dependsOn detekt if you want to run detekt on every build
repositories {

configurations {

task detekt(type: JavaExec) {
    main = "io.gitlab.arturbosch.detekt.cli.Main"
    classpath = configurations.detekt
    def input = "$projectDir"
    def config = "$projectDir/detekt.yml"
    def filters = ".*test.*"
    def rulesets = ""
    def params = [ '-i', input, '-c', config, '-f', filters, '-r', rulesets]

dependencies {
    detekt 'io.gitlab.arturbosch.detekt:detekt-cli:1.0.0.[version]'
    detekt 'io.gitlab.arturbosch.detekt:detekt-formatting:1.0.0.[version]'

Attention Android Developers! the dependencies section must be at the bottom, after the repository, configurations and task sections!

Using detekt in Maven Projects

  1. Add following lines to your pom.xml.
  2. Run mvn verify (when using the verify phase as I did here)
                    <!-- This can be run separately with mvn antrun:run@detekt -->
                        <target name="detekt">
                            <java taskname="detekt" dir="${basedir}" fork="true" failonerror="true"
                                  classname="io.gitlab.arturbosch.detekt.cli.Main" classpathref="maven.plugin.classpath">
                                <arg value="-p"/>
                                <arg value="${basedir}/src"/>
                                <arg value="-f"/>
                                <arg value=".*test.*"/>
                                <arg value="--useTabs"/>

    <name>arturbosch-code-analysis (for detekt)</name>


Currently there are seven rule sets which are used per default when running the cli.

  • complexity - has rules to detect LongMethod, LongParameterList, LargeClass, ComplexMethod ... smells
  • code-smell - other rules which can be classified as code smells but do not fit into the complexity category
  • style - detects wildcard imports and naming violations
  • comments - has rules to detect missing KDoc over public members and unnecessary KDoc over private members
  • exceptions - too general exceptions are used in throw and catch statements like RuntimeException, Error or Throwable
  • empty - finds empty block statements
  • potential-bugs - code is structured in a way it can lead to bugs like 'only equals but not hashcode is implemented' or explicit garbage collection calls
  • performance - finds potential performance issues
Additional RuleSets
  • formatting: detects indentation, spacing problems and optional semicolons in code

As of milestone six, the formatting rule set is shipped as an standalone plugin which must be linked to a detekt run through the --rules "path/to/jar" parameter or via gradle/maven classpath setup.

RuleSet Configuration

To turn off specific rules/rule sets or change threshold values for certain rules a yaml configuration file can be used. Export the default config with the --generate-config flag or copy and modify the detekt-cli/src/main/resources/default-detekt-config.yml for your needs.

  active: true
    threshold: 20
    threshold: 5
    threshold: 150
    threshold: 10
    threshold: 10
    threshold: 3

  active: true
    active: true
    active: true
    active: true
    variablePattern: '^(_)?[a-z$][a-zA-Z$0-9]*$'
    constantPattern: '^([A-Z_]*|serialVersionUID)$'
    methodPattern: '^[a-z$][a-zA-Z$0-9]*$'
    classPattern: '[A-Z$][a-zA-Z$]*'
    enumEntryPattern: '^[A-Z$][a-zA-Z_$]*$'
    active: true
    maxLineLength: 120
    excludePackageStatements: false
    excludeImportStatements: false

  active: true
    active: true
    active: true
    active: false
    active: false

  active: true
    active: true
    active: true
    active: true
    active: false

  active: true
    active: true

  active: true

  active: true
autoCorrect: true
    autoCorrect: true

Every rule of the default rule sets can be turned off. Thresholded code-smell rules can have an additional field threshold. Formatting rules can be configured to autoCorrect the style mistake.

Active keyword is only needed if you want to turn off the rule. Active on the rule set level turn off whole rule set. autoCorrect on the top level must be set to true or else all configured formatting rules are ignored. This is done to prevent you from changing your project files if your not 100% sure about it.

Suppress code smell rules

detekt supports the Java (@SuppressWarnings) and Kotlin (@Suppress) style suppression. If both annotations are present, only Kotlin's annotation is used! To suppress a rule, the id of the rule must be written inside the values field of the annotation e.g. @Suppress("LongMethod", "LongParameterList", ...)

Configure build failure thresholds

detekt now can throw a BuildFailure(Exception) and let the build fail with following config parameters:

  warningThreshold: 5 // Five weighted findings 
  failThreshold: 10 // Ten weighted smells to fail the build
    complexity: 2 // Whole complexity rule should add two for each finding.
    LongParameterList: 1 // The specific rule should not add two.
    comments: 0 // Comments should not fail the build at all?!

Every rule and rule set can be attached with an integer value which is the weight of the finding. For example: If you have 5 findings of the category code-smell, then your failThreshold of 10 is reached as 5 x 2 = 10.

The formula for weights: RuleID > RuleSetID > 1. Only integer values are supported.

Extending detekt

Custom RuleSets

detekt uses a ServiceLoader to collect all instances of RuleSetProvider-interfaces. So it is possible to define rules/rule sets and enhance detekt with your own flavor. Attention: You need a resources/META-INF/services/io.gitlab.arturbosch.detekt.api.RuleSetProvider file which has as content the fully qualified name of your RuleSetProvider e.g. io.gitlab.arturbosch.detekt.sampleruleset.SampleProvider.

The easiest way to define an rule set is to clone the provided detekt-sample-ruleset project.

Own rules have to extend the abstract Rule class and override the visitXXX functions from the AST. A RuleSetProvider must be implemented which declares a RuleSet in the instance method. To allow your rule to be configurable, pass it a Config object from within your rule set provider. You can also specify a Severity type for your rule.

Example of a custom rule:

class TooManyFunctions : Rule("TooManyFunctions") {

    private var amount: Int = 0

    override fun visitFile(file: PsiFile) {
        if (amount > 10) {
            addFindings(CodeSmell(id, Entity.from(file)))

    override fun visitNamedFunction(function: KtNamedFunction) {


Example of a much preciser rule in terms of more specific CodeSmell constructor and Rule attributes:

class TooManyFunctions2(config: Config) : Rule("TooManyFunctionsTwo", Severity.Maintainability, config) {

    private var amount: Int = 0

    override fun visitFile(file: PsiFile) {
        if (amount > 10) {
                    id = id, entity = Entity.from(file),
                    description = "Too many functions can make the maintainability of a file more costly",
                    metrics = listOf(Metric(type = "SIZE", value = amount, threshold = 10)),
                    references = listOf())

    override fun visitNamedFunction(function: KtNamedFunction) {


If you want your rule to be configurable, write down your properties inside the detekt.yml file and use the withConfig function:

    MyMetric: 5
    threshold: 10
    active: false

By specifying the rule set and rule ids, detekt will use the sub configuration of MyRule:

val threshold = withConfig { valueOrDefault("threshold") { threshold } }


If your using maven to build rule sets or use detekt as a dependency, you have to run the additional task install

Custom Processors


Custom Reports

detekt allows you to extend the console output and to create custom output formats.

For example if you do not like the default printing of findings, we can ... TODO

Testing your rules

To test your rules you need a KtFile object and use it's visit method. There are two predefined methods to help obtaining a KtFile:

  • compileContentForTest(content: String): KtFile
  • compileForTest(path: Path): KtFile

New with M3 there is a special detekt-test module, which specifies above two methods but also Rule extension functions that allow allow to skip compilation, ktFile and visit procedures.

  • Rule.lint(StringContent/Path) returns just the findings for given content
  • Rule.format(StringContent/Path) returns just the new modified content for given content


As of M11 the detektIdeaFormat-task makes the formatting rule set obsolete to a big part. It is highly advised to use that gradle task for indentation and spacing. Rules like removing OptionalUnit and converting to ExpressionBodySyntax for single return statements can still be useful.

KtLint was first to support auto correct formatting according to the kotlin coding conventions. In detekt I made an effort to port over all available formatting rules to detect style violations and auto correct them.

Following configuration I use to check the style for detekt. If your like me who prefer tabs over spaces, use useTabs in the rule set level to turn off indentation check for spaces (or simple turn off Indentation rule).

  active: true
  useTabs: true
    active: false
    indentSize: 4
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: true
    autoCorrect: true
    active: false
    autoCorrect: true
    active: false
    autoCorrect: true

Code Smell baseline and ignore list

Specify a report output with --output parameter and specify its format with --output-format. Now you can generate a report which holds all findings of current analysis.

With --baseline you generate a baseline.xml where code smells are white- or blacklisted.

  <Blacklist timestamp="1483388204705">
    <ID>CatchRuntimeException:Junk.kt$e: RuntimeException</ID>
  <Whitelist timestamp="1496432564542">
    <ID>NestedBlockDepth:Indentation.kt$Indentation$override fun procedure(node: ASTNode)</ID>
    <ID>ComplexCondition:SpacingAroundOperator.kt$SpacingAroundOperator$tokenSet.contains(node.elementType) &amp;&amp; node is LeafPsiElement &amp;&amp; !node.isPartOf(KtPrefixExpression::class) &amp;&amp; // not unary !node.isPartOf(KtTypeParameterList::class) &amp;&amp; // fun &lt;T&gt;fn(): T {} !node.isPartOf(KtTypeArgumentList::class) &amp;&amp; // C&lt;T&gt; !node.isPartOf(KtValueArgument::class) &amp;&amp; // fn(*array) !node.isPartOf(KtImportDirective::class) &amp;&amp; // import * !node.isPartOf(KtSuperExpression::class)</ID>
    <ID>ComplexMethod:DetektExtension.kt$DetektExtension$fun convertToArguments(): MutableList&lt;String&gt;</ID>

The intention of a whitelist is that only new code smells are printed on further analysis. The blacklist can be used to write down false positive detections. The ID node must be build of <RuleID>:<Signature>. Both values can be found inside the report file.



Android 开发经验分享
Android 开发经验分享