Native Datalevin

Datalevin can be compiled as a GraalVM native image.

Build Native Datalevin

Assuming that you can build JVM Datalevin already, i.e. you have JDK and lein, here are the steps to build native Datalevin binary on your platform.

Linux/MacOS

1. Install GraalVM
2. Intsall GraalVM native image
3. Run script/compile.

If the compilation is successful, two binaries will appear in this directory: dtlv and dtlv-test0. The former is the Datalevin command line shell, and the latter runs core Datalevin tests in native mode. These are standalone executable that you can immediately run.

Windows

Same as the above, except to run script/compile.bat instead in step 3.

If the build is successful, you will get dtlv.exe and `dtlv-test.exe**.

Compiling Datalevin Dependency to Native Image

If your application depends on Datalevin library, and you want to compile your application into a GraalVM native image, you need to use org.clojars.huahaiy/datalevin-native as the dependency, instead of datalevin/datalevin, because org.clojars.huahaiy/datalevin-native includes some GraalVM specific native code that will fail to compile in regular JVM. It also includes pre-compiled native dependencies for these platforms:

• Linux x86-64 (Ubuntu 20.04 and 22.04)
• MacOS arm64 (i.e. Apple Silicon M series)
• MacOS x86-64
• Windows x86-64

First build an uberjar of your application, then compile it with native-image command.

Like all Clojure applications, class initialization needs to be done at native image build time, i.e. add --features=InitAtBuildTimeFeature to native-image command.

During the native image build time, our class initialization code extracts native libraries from the org.clojars.huahaiy/datalevin-native jar and put them in the GraalVM's default CLibraryPath for the platform (e.g. ${GRAALVM_HOME}/lib/svm/clibraries/linux-amd64/). The files will be deleted upon build completion.

If you are uncomfortable with writing to the default location or lack the write permission for that directory, you can set an environment variable DTLV_LIB_EXTRACT_DIR in the shell before the native image build, and the native libraries will then be put there instead. If so, you must also add -H:CLibraryPath=${DTLV_LIB_EXTRACT_DIR} option to native-image command. The directory referred to by the environment variable must exist and is writable.

You also need to have these environment variables set before native image compilation:

export DTLV_COMPILE_NATIVE=true
export USE_NATIVE_IMAGE_JAVA_PLATFORM_MODULE_SYSTEM=false

Finally, look at our script/compile file as an example.

For CI/CD on various platforms, you may want to consult our Github Action (for Linux/MacOS AMD64), Appveoyor (for Windows), and Cirrus CI(for Apple Silicon) yaml files for examples.

Native Datalevin libraries are static, so your final application is standalone and requires only libc dependency on the OS to run.