This release emphasized consistency, new features, and improvements for the shared modules feature. It is compatible with Python 2.7, 3.3, 3.4, 3.5, and 3.6. Python 2.6 and 3.2 are no longer supported. Java 1.6 is no longer supported.
It is now possible to share memory between Python and Java using direct java.nio.Buffers and numpy ndarrays. The memory must be allocated by Java using ByteBuffer.allocateDirect(). The direct buffer must be placed in a jep.DirectNDArray before passing it to Python. Jep will convert the DirectNDArray to a numpy ndarray referencing the same memory. If the numpy ndarray is passed back to Java then the original DirectNDArray is passed back and the memory remains shared. This provides faster data transfer between Java and Python since there is no need to copy the memory. This can also simplify any use cases that frequently transfer the same data between Java and Python.
Strings that represent the full range of unicode values will now be passed correctly between Java and Python. Previous versions of Jep had minor inconsistencies for characters outside the range of US-ASCII and major problems transforming characters outside of the Basic Multilingual Plane. The conversion was also optimized to avoid unnecessary intermediary charset encodings. Python 3 versions before 3.3 use a more complex unicode object than newer versions of Python 3 so these versions are no longer supported. Unicode objects are uncommon in Python 2 so Java Strings will be converted to UTF-8 encoded Python string objects.
PyJAutoCloseable has been created to enable developers to more closely match Python syntax when working with Java AutoCloseables. The new type extends PyJObject to retain all previous functionality. This enables Python's "with" keyword and ContextManagers to be used similarly to Java 7's try-with-resources. For example:
from java.io import FileWriter
with FileWriter("/tmp/test.txt") as f:
f.write("abc")
Since every Java Object has an intrinsic lock associated with it, you can now synchronize on Java Objects from within Python. Every PyJObject has a synchronized() method attached to it to enter the lock. Developers should always use Python's "with" keyword to ensure the lock is released when the block of code is exited. The Java code
synchronized(obj) {
// code here
}
can be expressed in Python as
with obj.synchronized():
// code here
where obj is a PyJObject.
A Python function (or any other callable) can be passed to a Java method which takes a functional interface (such as java.lang.Runnable). Similar to Java 8, the type is not required to have the FunctionalInterface annotation to be treated as a functional interface. Although the concept of functional interfaces was not introduced until Java 8, Jep will allow the conversion on all supported Java versions. Similar to other Jep Proxy objects, the new Java Object can only be used on the thread where it was created. For Example:
def printOne():
print(1)
from java.util.concurrent import Executors
javaCallable = Executors.callable(printOne)
javaCallable.call()
Contributed by Techcable.