构建最终原生二进制文件

By default, a Kotlin/Native target is compiled down to a *.klib library artifact, which can be consumed by Kotlin/Native itself as a dependency but cannot be executed or used as a native library.

To declare final native binaries such as executables or shared libraries, use the binaries property of a native target. This property represents a collection of native binaries built for this target in addition to the default *.klib artifact and provides a set of methods for declaring and configuring them.

The kotlin-multiplatform plugin doesn’t create any production binaries by default. The only binary available by default is a debug test executable that lets you run unit tests from the test compilation.

构建最终原生二进制文件 - 图1

Binaries produced by the Kotlin/Native compiler can include third-party code, data, or derived work. This means if you distribute a Kotlin/Native-compiled final binary, you should always include necessary license files into your binary distribution.

声明二进制文件

Use the following factory methods to declare elements of the binaries collection.

Factory methodBinary kindAvailable for
executableProduct executableAll native targets
testTest executableAll native targets
sharedLibShared native libraryAll native targets, except for WebAssembly
staticLibStatic native libraryAll native targets, except for WebAssembly
frameworkObjective-C frameworkmacOS, iOS, watchOS, and tvOS targets only

The simplest version doesn’t require any additional parameters and creates one binary for each build type. Currently, two build types are available:

  • DEBUG – produces a non-optimized binary with debug information
  • RELEASE – produces an optimized binary without debug information

The following snippet creates two executable binaries, debug and release:

  1. kotlin {
  2. linuxX64 { // Define your target instead.
  3. binaries {
  4. executable {
  5. // Binary configuration.
  6. }
  7. }
  8. }
  9. }

You can drop the lambda if there is no need for additional configuration:

  1. binaries {
  2. executable()
  3. }

You can specify for which build types to create binaries. In the following example, only the debug executable is created:

【Kotlin】

  1. binaries {
  2. executable(listOf(DEBUG)) {
  3. // Binary configuration.
  4. }
  5. }

【Groovy】

  1. binaries {
  2. executable([DEBUG]) {
  3. // Binary configuration.
  4. }
  5. }

You can also declare binaries with custom names:

【Kotlin】

  1. binaries {
  2. executable("foo", listOf(DEBUG)) {
  3. // Binary configuration.
  4. }
  5. // It's possible to drop the list of build types
  6. // (in this case, all the available build types will be used).
  7. executable("bar") {
  8. // Binary configuration.
  9. }
  10. }

【Groovy】

  1. binaries {
  2. executable('foo', [DEBUG]) {
  3. // Binary configuration.
  4. }
  5. // It's possible to drop the list of build types
  6. // (in this case, all the available build types will be used).
  7. executable('bar') {
  8. // Binary configuration.
  9. }
  10. }

The first argument sets a name prefix, which is the default name for the binary file. For example, for Windows the code produces the files foo.exe and bar.exe. You can also use the name prefix to access the binary in the build script.

访问二进制文件

You can access binaries to configure them or get their properties (for example, the path to an output file).

You can get a binary by its unique name. This name is based on the name prefix (if it is specified), build type, and binary kind following the pattern: <optional-name-prefix><build-type><binary-kind>, for example, releaseFramework or testDebugExecutable.

Static and shared libraries have the suffixes static and shared respectively, for example, fooDebugStatic or barReleaseShared.

构建最终原生二进制文件 - 图2

【Kotlin】

  1. // Fails if there is no such binary.
  2. binaries["fooDebugExecutable"]
  3. binaries.getByName("fooDebugExecutable")
  4. // Returns null if there is no such binary.
  5. binaries.findByName("fooDebugExecutable")

【Groovy】

  1. // Fails if there is no such binary.
  2. binaries['fooDebugExecutable']
  3. binaries.fooDebugExecutable
  4. binaries.getByName('fooDebugExecutable')
  5. // Returns null if there is no such binary.
  6. binaries.findByName('fooDebugExecutable')

Alternatively, you can access a binary by its name prefix and build type using typed getters.

【Kotlin】

  1. // Fails if there is no such binary.
  2. binaries.getExecutable("foo", DEBUG)
  3. binaries.getExecutable(DEBUG) // Skip the first argument if the name prefix isn't set.
  4. binaries.getExecutable("bar", "DEBUG") // You also can use a string for build type.
  5. // Similar getters are available for other binary kinds:
  6. // getFramework, getStaticLib and getSharedLib.
  7. // Returns null if there is no such binary.
  8. binaries.findExecutable("foo", DEBUG)
  9. // Similar getters are available for other binary kinds:
  10. // findFramework, findStaticLib and findSharedLib.

【Groovy】

  1. // Fails if there is no such binary.
  2. binaries.getExecutable('foo', DEBUG)
  3. binaries.getExecutable(DEBUG) // Skip the first argument if the name prefix isn't set.
  4. binaries.getExecutable('bar', 'DEBUG') // You also can use a string for build type.
  5. // Similar getters are available for other binary kinds:
  6. // getFramework, getStaticLib and getSharedLib.
  7. // Returns null if there is no such binary.
  8. binaries.findExecutable('foo', DEBUG)
  9. // Similar getters are available for other binary kinds:
  10. // findFramework, findStaticLib and findSharedLib.

将依赖项导出到二进制文件

When building an Objective-C framework or a native library (shared or static), you may need to pack not just the classes of the current project, but also the classes of its dependencies. Specify which dependencies to export to a binary using the export method.

【Kotlin】

  1. kotlin {
  2. sourceSets {
  3. macosMain.dependencies {
  4. // Will be exported.
  5. api(project(":dependency"))
  6. api("org.example:exported-library:1.0")
  7. // Will not be exported.
  8. api("org.example:not-exported-library:1.0")
  9. }
  10. }
  11. macosX64("macos").binaries {
  12. framework {
  13. export(project(":dependency"))
  14. export("org.example:exported-library:1.0")
  15. }
  16. sharedLib {
  17. // It's possible to export different sets of dependencies to different binaries.
  18. export(project(':dependency'))
  19. }
  20. }
  21. }

【Groovy】

  1. kotlin {
  2. sourceSets {
  3. macosMain.dependencies {
  4. // Will be exported.
  5. api project(':dependency')
  6. api 'org.example:exported-library:1.0'
  7. // Will not be exported.
  8. api 'org.example:not-exported-library:1.0'
  9. }
  10. }
  11. macosX64("macos").binaries {
  12. framework {
  13. export project(':dependency')
  14. export 'org.example:exported-library:1.0'
  15. }
  16. sharedLib {
  17. // It's possible to export different sets of dependencies to different binaries.
  18. export project(':dependency')
  19. }
  20. }
  21. }

For example, you implement several modules in Kotlin and want to access them from Swift. Usage of several Kotlin/Native frameworks in a Swift application is limited, but you can create an umbrella framework and export all these modules to it.

You can export only api dependencies of the corresponding source set.

构建最终原生二进制文件 - 图3

When you export a dependency, it includes all of its API to the framework API. The compiler adds the code from this dependency to the framework, even if you use a small fraction of it. This disables dead code elimination for the exported dependency (and for its dependencies, to some extent).

By default, export works non-transitively. This means that if you export the library foo depending on the library bar, only methods of foo are added to the output framework.

You can change this behavior using the transitiveExport option. If set to true, the declarations of the library bar are exported as well.

It is not recommended to use transitiveExport: it adds all transitive dependencies of the exported dependencies to the framework. This could increase both compilation time and binary size.

In most cases, you don’t need to add all these dependencies to the framework API. Use export explicitly for the dependencies you need to directly access from your Swift or Objective-C code.

构建最终原生二进制文件 - 图4

【Kotlin】

  1. binaries {
  2. framework {
  3. export(project(":dependency"))
  4. // Export transitively.
  5. transitiveExport = true
  6. }
  7. }

【Groovy】

  1. binaries {
  2. framework {
  3. export project(':dependency')
  4. // Export transitively.
  5. transitiveExport = true
  6. }
  7. }

构建 universal frameworks

By default, an Objective-C framework produced by Kotlin/Native supports only one platform. However, you can merge such frameworks into a single universal (fat) binary using the lipo tool. This operation especially makes sense for 32-bit and 64-bit iOS frameworks. In this case, you can use the resulting universal framework on both 32-bit and 64-bit devices.

The fat framework must have the same base name as the initial frameworks. Otherwise, you’ll get an error.

构建最终原生二进制文件 - 图5

【Kotlin】

  1. import org.jetbrains.kotlin.gradle.tasks.FatFrameworkTask
  2. kotlin {
  3. // Create and configure the targets.
  4. val ios32 = watchosArm32("watchos32")
  5. val ios64 = watchosArm64("watchos64")
  6. configure(listOf(watchos32, watchos64)) {
  7. binaries.framework {
  8. baseName = "my_framework"
  9. }
  10. }
  11. // Create a task to build a fat framework.
  12. tasks.register<FatFrameworkTask>("debugFatFramework") {
  13. // The fat framework must have the same base name as the initial frameworks.
  14. baseName = "my_framework"
  15. // The default destination directory is "<build directory>/fat-framework".
  16. destinationDir = buildDir.resolve("fat-framework/debug")
  17. // Specify the frameworks to be merged.
  18. from(
  19. ios32.binaries.getFramework("DEBUG"),
  20. ios64.binaries.getFramework("DEBUG")
  21. )
  22. }
  23. }

【Groovy】

  1. import org.jetbrains.kotlin.gradle.tasks.FatFrameworkTask
  2. kotlin {
  3. // Create and configure the targets.
  4. targets {
  5. watchosArm32("watchos32")
  6. watchosArm64("watchos64")
  7. configure([watchos32, watchos64]) {
  8. binaries.framework {
  9. baseName = "my_framework"
  10. }
  11. }
  12. }
  13. // Create a task building a fat framework.
  14. tasks.register("debugFatFramework", FatFrameworkTask) {
  15. // The fat framework must have the same base name as the initial frameworks.
  16. baseName = "my_framework"
  17. // The default destination directory is "<build directory>/fat-framework".
  18. destinationDir = file("$buildDir/fat-framework/debug")
  19. // Specify the frameworks to be merged.
  20. from(
  21. targets.ios32.binaries.getFramework("DEBUG"),
  22. targets.ios64.binaries.getFramework("DEBUG")
  23. )
  24. }
  25. }

Build XCFrameworks

All Kotlin Multiplatform projects can use XCFrameworks as an output to gather logic for all the target platforms and architectures in a single bundle. Unlike universal (fat) frameworks, you don’t need to remove all unnecessary architectures before publishing the application to the App Store.

【Kotlin】

  1. import org.jetbrains.kotlin.gradle.plugin.mpp.apple.XCFramework
  2. plugins {
  3. kotlin("multiplatform")
  4. }
  5. kotlin {
  6. val xcf = XCFramework()
  7. val iosTargets = listOf(iosX64(), iosArm64(), iosSimulatorArm64())
  8. iosTargets.forEach {
  9. it.binaries.framework {
  10. baseName = "shared"
  11. xcf.add(this)
  12. }
  13. }
  14. }

【Groovy】

  1. import org.jetbrains.kotlin.gradle.plugin.mpp.apple.XCFrameworkConfig
  2. plugins {
  3. id 'org.jetbrains.kotlin.multiplatform'
  4. }
  5. kotlin {
  6. def xcf = new XCFrameworkConfig(project)
  7. def iosTargets = [iosX64(), iosArm64(), iosSimulatorArm64()]
  8. iosTargets.forEach {
  9. it.binaries.framework {
  10. baseName = 'shared'
  11. xcf.add(it)
  12. }
  13. }
  14. }

When you declare XCFrameworks, Kotlin Gradle plugin will register three Gradle tasks:

  • assembleXCFramework
  • assembleDebugXCFramework (additionally debug artifact that contains dSYMs)
  • assembleReleaseXCFramework

If you’re using CocoaPods integration in your projects, you can build XCFrameworks with the Kotlin CocoaPods Gradle plugin. It includes the following tasks that build XCFrameworks with all the registered targets and generate podspec files:

  • podPublishReleaseXCFramework, which generates a release XCFramework along with a podspec file.
  • podPublishDebugXCFramework, which generates a debug XCFramework along with a podspec file.
  • podPublishXCFramework, which generates both debug and release XCFrameworks along with a podspec file.

This can help you distribute shared parts of your project separately from mobile apps through CocoaPods. You can also use XCFrameworks for publishing to private or public podspec repositories.

Publishing Kotlin frameworks to public repositories is not recommended if those frameworks are built for different versions of Kotlin. Doing so might lead to conflicts in the end-users’ projects.

构建最终原生二进制文件 - 图6

Customize the Info.plist file

When producing a framework, the Kotlin/Native compiler generates the information property list file, Info.plist. You can customize its properties with the corresponding binary option:

PropertyBinary option
CFBundleIdentifierbundleId
CFBundleShortVersionStringbundleShortVersionString
CFBundleVersionbundleVersion

To enable the feature, pass the -Xbinary=$option=$value compiler flag or set the binaryOption("option", "value") Gradle DSL for the specific framework:

  1. binaries {
  2. framework {
  3. binaryOption("bundleId", "com.example.app")
  4. binaryOption("bundleVersion", "2")
  5. }
  6. }