Posted by Ting-Yuan Huang, Software Engineer and Jiaxiang Chen, Software
Engineer
Kotlin Symbol Processing (KSP),
our new tool for building lightweight compiler plugins in Kotlin, is now
stable! KSP offers similar functionality to the Kotlin Annotation Processing
Tool (KAPT),
however it’s up to 2x faster, offers direct access to Kotlin language
constructs, and offers support for multiplatform targets.
Over the past few months, KSP has gone through 32 releases with over 162 bugs
reported from the community and fixed by our team. If you were waiting to
adopt it, now is the time to check it out.
Why we built KSP
On the Android team, we regularly ask developers: what are your biggest
frustrations with writing apps today? One of the top issues that comes up
repeatedly is build speed. Over the years we’ve been making steady
improvements to the Android build toolchain, and today we’re excited to add to
those improvements with KSP. KSP is the next generation of annotation
processing in Kotlin: it will dramatically improve build speed for Kotlin
developers, and unlike KAPT, it offers support for Kotlin/Native and
Kotlin/JS.
“Adding KSP support to Room improved the compilation speed and also enabled
Room to better understand Kotlin code, such as the nullability of generics
that was not possible with KAPT. It also unlocks new possibilities like
generating Kotlin code, which will allow Room to have a better Kotlin user
experience in the future.” – Yigit Boyar, Software Engineer, Android
Why is KSP faster?
The Kotlin Annotation Processing Tool (KAPT) works with Java’s annotation processing infrastructure to make most Java
language annotation processors work in Kotlin out of the box. To do this, KAPT
compiles Kotlin code into Java stubs that retain information that Java
annotation processors care about. Creating these stubs is costly though, and
means the compiler must resolve all the symbols in your program multiple times
(once to generate stubs, and then again to do the actual compilation).
KSP moves away from the stub generation model by working as a Kotlin compiler
plugin — it allows annotation processors to read and analyze source programs
and resources directly in Kotlin instead of requiring you to depend on the
Java annotation processing infrastructure. This both dramatically improves
build speed (up to 2x faster for Room’s
Kotlin test app) and means that KSP can be used for non-Android and non-JVM environments
like Kotlin/Native and Kotlin/JS.
How to get started
To start using KSP, download the
KSP playground project
from GitHub, which shows how to use KSP both as an annotation processor and as
a consuming app/library:
-
Annotation processor: A toy
test-processorlibrary that implements the builder pattern as a
KSP processor -
Consuming library: A
workloaddirectory that
shows how to use the builder processor in a real-world Kotlin project
If you’re an app developer, check out the
list of supported libraries
and the
quickstart
guide for moving a module over from KAPT to KSP.
Using Moshi or Room with KSP
If you’re using Moshi or Room in your project, you can already try out KSP by
making a quick fix to your module’s build file. For example, to use the KSP
version of Room in a Gradle module you can simply replace the KAPT plugin with
KSP and swap out the KSP dependency:
apply plugin: 'com.google.devtools.ksp'
dependencies {
...
implementation "androidx.room:room-runtime:$room_version"
kapt "androidx.room:room-compiler:$room_version"
ksp "androidx.room:room-compiler:$room_version"
}
Check out the
Room release notes
for more info.
Conclusion
With the 1.0 release of KSP you will start to see improved build times for
your Kotlin projects as you migrate away from libraries based on KAPT. We have
also updated a number of Android specific libraries which are ready for you to
try today and offer significant performance improvements.
