Java bridge packages#

Corese is a stack of Java libraries that implement the standards of the Semantic Web. To run and access Java libraries we tested two off-the-shelf packages:

py4j - this package establishes a bridge between Python and Java using sockets (IPC). The

corese-python Java library implements the listener and wrapper for corese-core library and is built with gradle specifically for pycorese.

jpype - this package utilizes a single shared memory space by embedding the JVM directly into the Python process. The actual corese-core Java library is downloaded from the Maven repository.

At first we tried both packages to decide which one is better for our purposes. We found that both py4j and jpype have their pros and cons. See the comparison table below:

Compare Python- Java bridge packages#
	JPype	Py4J
Core Technology	JNI (direct JVM embedding)	Reflection & Socket IPC
Communication	Shared memory (in-process)	Socket based (out-of-process)
Performance	High (no serialization)	Moderate (due to IPC overhead)
Setup	Requires JVM in-process	Requires a GatewayServer
Use case	High-performance scenarios	Lightweight, flexible usage
Crash Tolearnce	Low: JVM crash affects Python (shared process) and cannot be restarted on its own	High: JVM and Python are separate processes, so a JVM crash won’t take down Python.
Distribution	Download `corese-core-{version}.jar`	Build `corese-python-{version}.jar`

Although jpype is more performant and does not require an extra wrapper library, we decided to use py4j as a default because it is more crash tolerant. The main reason is that jpype requires the JVM to be embedded in the Python process and prohibits the restart of this subprocess in case of a crash (by design). This could be a critical issue for long-running applications or services.

However we decided to leave both bridges in the package and let the user choose the one that fits their needs better.

Both Java Libraries are installed with the package and can be used interchangeably. See the User Guide.