I am wondering if it is possible to call Python functions from Java code using Jython, or is it only for calling Java code from Python?
Jython: Python for the Java Platform - http://www.jython.org/index.html
You can easily call python functions from Java code with Jython. That is as long as your python code itself runs under jython, i.e. doesn't use some c-extensions that aren't supported.
If that works for you, it's certainly the simplest solution you can get. Otherwise you can use org.python.util.PythonInterpreter from the new Java6 interpreter support.
A simple example from the top of my head - but should work I hope: (no error checking done for brevity)
PythonInterpreter interpreter = new PythonInterpreter();
interpreter.exec("import sys\nsys.path.append('pathToModules if they are not there by default')\nimport yourModule");
// execute a function that takes a string and returns a string
PyObject someFunc = interpreter.get("funcName");
PyObject result = someFunc.__call__(new PyString("Test!"));
String realResult = (String) result.__tojava__(String.class);
As of 2021, Jython does not support Python 3.x
I think there are some important things to consider first with how strong you wish to have the linking between java and python.
Firstly Do you only want to call functions or do you actually want python code to change the data in your java objects? This is very important. If you only want to call some python code with or without arguments, then that is not very difficult. If your arguments are primitives it makes it even more easy. However if you want to have java class implement member functions in python, which change the data of the java object, then this is not so easy or straight forward.
Secondly are we talking cpython or will jython do? I would say cpython is where its at! I would advocate this is why python is so kool! Having such high abstractions however access to c,c++ when needed. Imagine if you could have that in java. This question is not even worth asking if jython is ok because then it is easy anyway.
So I have played with the following methods, and listed them from easy to difficult:
Java to Jython
Advantages: Trivially easy. Have actual references to java objects
Disadvantages: No CPython, Extremely Slow!
Jython from java is so easy, and if this is really enough then great. However it is very slow and no cpython! Is life worth living without cpython I don't think so! You can easily have python code implementing your member functions for you java objects.
Java to Jython to CPython via Pyro
Pyro is the remote object module for python. You have some object on a cpython interpreter, and you can send it objects which are transferred via serialization and it can also return objects via this method. Note that if you send a serialized python object from jython and then call some functions which change the data in its members, then you will not see those changes in java. You just need to remember to send back the data which you want from pyro. This I believe is the easiest way to get to cpython! You do not need any jni or jna or swig or .... You don't need to know any c, or c++. kool huh?
Advantages: Access to cpython, not as difficult as following methods
Disadvantages: Cannot change the member data of java objects directly from python. Is somewhat indirect, (jython is middle man).
Java to C/C++ via JNI/JNA/SWIG to Python via Embedded interpreter (maybe using BOOST Libraries?)
OMG this method is not for the faint of heart. And I can tell you it has taken me very long to achieve this in with a decent method. Main reason you would want to do this is so that you can run cpython code which as full rein over you java object. There are major major things to consider before deciding to try and bread java (which is like a chimp) with python (which is like a horse). Firstly if you crash the interpreter that's lights out for you program! And don't get me started on concurrency issues! In addition, there is allot allot of boiler, I believe I have found the best configuration to minimize this boiler but still it is allot! So how to go about this:
Consider that C++ is your middle man, your objects are actually c++ objects! Good that you know that now. Just write your object as if your program as in cpp not java, with the data you want to access from both worlds. Then you can use the wrapper generator called swig (http://www.swig.org/Doc1.3/Java.html) to make this accessible to java and compile a dll which you call System.load(dll name here) in java. Get this working first, then move on to the hard part!
To get to python you need to embed an interpreter. Firstly I suggest doing some hello interpreter programs or this tutorial Embedding python in C/C. Once you have that working, its time to make the horse and the monkey dance! You can send you c++ object to python via [boost][3] . I know I have not given you the fish, merely told you where to find the fish. Some pointers to note for this when compiling.
When you compile boost you will need to compile a shared library. And you need to include and link to the stuff you need from jdk, ie jawt.lib, jvm.lib, (you will also need the client jvm.dll in your path when launching the application) As well as the python27.lib or whatever and the boost_python-vc100-mt-1_55.lib.
Then include Python/include, jdk/include, boost and only use shared libraries (dlls) otherwise boost has a teary. And yeah full on I know. There are so many ways in which this can go sour. So make sure you get each thing done block by block. Then put them together.
It's not smart to have python code inside java. Wrap your python code with flask or another web framework to make it a microservice. This makes your java program able to call this microservice (e.g. via REST).
This approach is simple and it will save you tons of issues. And the codes are loosely coupled so they are scalable.
Updated on Mar 24th 2020:
According to #stx's comment, the above approach is not suitable for massive data transfer between client and server.
Here is another approach I recommended:
Connecting Python and Java with Rust(C/C++ also ok).
https://medium.com/#shmulikamar/https-medium-com-shmulikamar-connecting-python-and-java-with-rust-11c256a1dfb0
Several of the answers mention that you can use JNI or JNA to access cpython but I would not recommend starting from scratch because there are already open source libraries for accessing cpython from java. For example:
JEP
JPY
GraalVM is a good choice. I've done Java+Javascript combination with GraalVM for microservice design (Java with Javascript reflection). They recently added support for python, I'd give it a try especially with how big its community has grown over the years.
UPDATE June 2021
https://www.graalvm.org/reference-manual/python/ says
GraalVM provides a Python 3.8 compliant runtime. A primary goal of the GraalVM Python runtime is to support SciPy and its constituent libraries, as well as to work with other data science and machine learning libraries from the rich Python ecosystem. At this point, the Python runtime is made available for experimentation and curious end-users.
Here a library that lets you write your python scripts once and decide which integration method (Jython, CPython/PyPy via Jep and Py4j) to use at runtime:
https://github.com/subes/invesdwin-context-python
Since each method has its own benefits/drawbacks as explained in the link.
(3 modules: jep, Py4J, jython )
It depends on what do you mean by python functions? if they were written in cpython you can not directly call them you will have to use JNI, but if they were written in Jython you can easily call them from java, as jython ultimately generates java byte code.
Now when I say written in cpython or jython it doesn't make much sense because python is python and most code will run on both implementations unless you are using specific libraries which relies on cpython or java.
see here how to use Python interpreter in Java.
Depending on your requirements, options like XML-RPC could be useful, which can be used to remotely call functions virtually in any language supporting the protocol.
Jython has some limitations:
There are a number of differences. First, Jython programs cannot use CPython
extension modules written in C. These modules usually have files with the
extension .so, .pyd or .dll. If you want to use such a module, you should look
for an equivalent written in pure Python or Java. Although it is technically
feasible to support such extensions - IronPython does so - there are no plans
to do so in Jython.
Distributing my Python scripts as JAR files with Jython?
you can simply call python scripts (or bash or Perl scripts) from Java using Runtime or ProcessBuilder and pass output back to Java:
Running a bash shell script in java
Running Command Line in Java
java runtime.getruntime() getting output from executing a command line program
You can call any language from java using Java Native Interface
This gives a pretty good overview over the current options. Some of which are named in other answers. Jython is not usable until they decide to not implement Python 3. Many of the other projects are coming from the python side and want to access java. But there are a few options still, to name something which has not been named yet: gRPC
I have similar requirement, I think best solution is thrift for me(also a rpc solution), just run test passed successfully right now, and can use thrift-generator to gen thrift file from java interface, then gen python files and java client files from the thrift file
Related
I have a Java app that needs to integrate with a 3rd party library. The library is written in Python, and I don't have any say over that. I'm trying to figure out the best way to integrate with it. I'm trying out JEPP (Java Embedded Python) - has anyone used that before? My other thought is to use JNI to communicate with the C bindings for Python.
Any thoughts on the best way to do this would be appreciated. Thanks.
Why not use Jython? The only downside I can immediately think of is if your library uses CPython native extensions.
EDIT: If you can use Jython now but think you may have problems with a later version of the library, I suggest you try to isolate the library from your app (e.g. some sort of adapter interface). Go with the simplest thing that works for the moment, then consider JNI/CPython/etc if and when you ever need to. There's little to be gained by going the (painful) JNI route unless you really have to.
Frankly most ways to somehow run Python directly from within JVM don't work. They are either not-quite-compatible (new release of your third party library can use python 2.6 features and will not work with Jython 2.5) or hacky (it will break with cryptic JVM stacktrace not really leading to solution).
My preferred way to integrate the two would use RPC. XML RPC is not a bad choice here, if you have moderate amounts of data. It is pretty well supported — Python has it in its standard library. Java libraries are also easy to find. Now depending on your setup either Java or Python part would be a server accepting connection from other language.
A less popular but worth considering alternative way to do RPCs is Google protobuffers, which have 2/3 of support for nice rpc. You just need to provide your transport layer. Not that much work and the convenience of writing is reasonable.
Another option is to write a C wrapper around that pieces of Python functionality that you need to expose to Java and use it via JVM native plugins. You can ease the pain by going with SWIG SWIG.
Essentially in your case it works like that:
Create a SWIG interface for all method calls from Java to C++.
Create C/C++ code that will receive your calls and internally call python interpreter with right params.
Convert response you get from python and send it via swig back to your Java code.
This solution is fairly complex, a bit of an overkill in most cases. Still it is worth doing if you (for some reason) cannot afford RPCs. RPC still would be my preferred choice, though.
Many years later, just to add an option which is more popular these days...
If you need CPython functionality, py4j is a good option. py4j has seen seen frequent updates in 2016 2017 2018 2019 2020 and has gained some popularity, because it is used e.g. by Apache Spark to achieve CPython interoperability.
The best solutions, is to use Python programs throw REST API. You define your services and call them. You perhaps need to learn some new modules. But you will be more flexible for futures changes.
Here a small list of use full modules for this purpose:
Python modules
Flask
Flask-SQLAlchemy
Flask-Restful
SQlite3
Jsonify
Java modules (for calling rest api)
Jersey or Apache CXF
You will need a small Learning curve, but later you will get more productivity and modularity and even elasticity...
My other thought is to use JNI to communicate with the C bindings for Python.
I like very much JNA:
JNA provides Java programs easy access to native shared libraries (DLLs on Windows) without writing anything but Java code—no JNI or native code is required. This functionality is comparable to Windows' Platform/Invoke and Python's ctypes. Access is dynamic at runtime without code generation.
My 0.02$ :)
You could use a messaging service like ActiveMQ. It has both Python and Java support. This way, you can leave the complicated JNI or C bindings as they are and deal solely with what I consider a simple interface. Moreover, when the library gets updated, you don't need to change much, if anything.
Have you considered running Jython on the Java VM?
I've investigated a similar setup with JNI. Maybe this will help if haven't seen it yet:
http://wiki.cacr.caltech.edu/danse/index.php/Communication_between_Java_and_Python
http://jpe.sourceforge.net/
These are some of the tools which make it easier to bridge the gap between Python and Java:
1.Jython
Python implemented in Java
2.JPype
Allows Python to run java commands
3.Jepp
Java embedded Python
4.JCC
a C++ code generator for calling Java from C++/Python
5.Javabridge
a package for running and interacting with the JVM from CPython
6.py4j
Allows Python to run java commands.
7.voc
Part of BeeWare suite. Converts python code to Java bytecode.
8.p2j
Converts Python code to Java. No longer developed.
If you can get your Python code to work in Jython, then you should be able to use that to call it from Java:
http://jython.sourceforge.net/cgi-bin/faqw.py?req=show&file=faq06.001.htp
I also think that run command line in the Java wouldn't by bad practice (stackoverflow question here).
Potentially share the data through some database.
I like the way to connect two apps via the bash pipe, but I have not practice in this, so I'm wondering how difficult is to write logic to handle this on both python/java sides.
Or other productive way could be to use the Remote Procedure Call (RPC) which supports procedural programming. Using RPC you can invokes methods in shared environments. As an example you can call a function in a remote machine from the local computer using RPC. We can define RPC as a communication type in distributed systems.
(mentioned above by Marcin)
Or, very naive way would to be to communicate by the common file. But for sake of simplicity and speed, my vote is to use the shared database x rest API x socket communication.
Also I like the XML RPC as Marcin wrote.
I would like to recommend to avoid any complication to run Python under JVM or C++ binding. Better to use today trends which are obviously web technologies.
As a shared database the MongoDB may be good solution or even better the Redis as in memory database.
Use pipes to communicate with python(a python script calls python library) via subprocesses, the whole process is like java<-> Pipes <-> py.
If JNI, you should be familiar with python's bindings(not recommended) and compiled files like *.so to work with java.
The whole process is like py -> c -> .so/.pyd -> JNI -> jar.
Here is a good practice of stdio.
https://github.com/JULIELab/java-stdio-ipc
I have a large programme written in C++ that needs to use a specific Java library. Ideally I would like to create an equivalent C++ library that wraps this existing Java code. As such I have been looking into the JNI invocation API.
Since I am not a very experienced programmer, and I am also inexperienced with JNI and multi-language programming, I would greatly appreciate some general pointers/tips/advice as to how to tackle this problem.
Things I would be especially interested to know:
Should each Java class in the Java lib map to to an associated C++
class? I.e. in my C++ library, will I have a class each invoking a
JVM for a particular Java class? Or will I have a singular JVM through which everything is accessed? What is the best way to do this and why?
What will be the basic process and architecture for doing this?
Are there any specific resources for creating a C++ lib from Java lib using the invocation API?
Thanks a lot!
I've done this before, but it's not for the faint-hearted especially if your interface between the 2 languages is hard. Debugging can also be a pain in this situation.
To answer your points:
You should start by deciding on what functionality from the Java library you need to access in your C++ program. Is it just a few tasks? Try making a very simple interface from C++ to Java in this case. Is it complicated? Then you're gonna have to start mapping Java classes to C++, and the more you need then the more work it's gonna be.
The end of q1 is sorta q2 really. Your C++ program will start a single JVM which will run as part of your program. When you make calls across the C++ data will be transferred into the JVM, and then the Java code executed, and then the return values transferred back. This incurs a performance cost so calling small functions like add(int,int) through JNI would be more expensive than just doing it in C++.
There's a lot of basic guides you can Google to get started. Just managing to start a basic JVM from C++ and making a call is actually a bit of work since you need to get the paths to the JVM libs correct or it doesn't work (unless they've improved this, it's been years since I tried). So you might want to check that out first before asking more specific questions about JNI and mapping functions.
An alternative option (which may or may-not be possible depending on your library and use-case) is to just write some kind of wrapper service around your library, actually in Java. And then send requests to it via JSON-HTTP or some messaging system.
An even-more alternative option, rewrite whatever the library is doing in C++.
You can use scapix::link::java C++ JNI library to generate C++ headers for any Java code, then easily access this Java code from C++. Here is an example:
#include <scapix/java_api/java/lang/System.h>
#include <scapix/java_api/java/util/Locale.h>
#include <scapix/java_api/java/text/DateFormatSymbols.h>
using namespace scapix::link::java;
using namespace scapix::java_api;
void test1()
{
// C++ objects are automatically converted to and from corresponding Java types.
// This works for any type supported by scapix::link::java::convert() interface,
// which supports many STL types and can be extended for your own types.
std::string version = java::lang::System::getProperty("java.version");
std::vector<std::string> languages = java::util::Locale::getISOLanguages();
std::vector<std::vector<std::string>> zone_strings = java::text::DateFormatSymbols::getInstance()->getZoneStrings();
std::map<std::string, std::string> properties = java::lang::System::getProperties();
}
A quick question that may seem out of the ordinary. (in reverse)
Instead of calling native code from an interpreted language; is there a way to compile Java or Python code to a .dll/.so and call the code from C/C++?
I'm willing to accept even answers such as manually spawning the interpreter or JVM and force it to read the .class/.py files. (is this a good solution?)
Thank you.
gcj can compile most Java source to native code (linked with a libgcj shared library) instead of to JVM bytecode.
There are a number of Python projects that are similar, like shedskin, but none as mature or active.
Cython is similar, but not quite the same—it compiles modules written in a Python-like language into native C extension modules for CPython. But if you put that together with embedding Python in a C app, it gives you most of what you want. But you are still running a Python interpreter loop to tie all those compiled-to-C functions together.
You can also do the same thing with Java—embed the JVM into your app, use gcj to compile any parts you want to native code, while compiling other parts to bytecode, and using JNI to communicate between them.
And of course you can use Jython to embed your Python code into the JVM, which you can embed into your C program, and because you can use JNI directly from Jython any pair of the three languages can effectively talk to each other without going through the third.
The idea of spawning a JVM or a CPython interpreter as a subprocess, which I think you were suggesting in your question, also works just fine. However, the only interface you will have to it in that case will be the child process's stdin/stdout/stderr (or any pipes or sockets you create manually), which isn't as flexible as being able to call methods directly on objects, etc. (Then again, sometimes that extra indirection can be a good thing, forcing you to define a cleanly-separated API between your components.)
You can embed a Python interpreter in your C/C++ program.
http://docs.python.org/2/extending/embedding.html
With Java your probably want the Java Native Interface (which works in both directions).
http://en.wikipedia.org/wiki/Java_Native_Interface
You can also look into Lua, while not as widely used as a lot of other scripting languages, it was meant to be embedded easily into executables. It's relatively small and fast. Just another option. If you want to call other languages from your c/c++ look into SWIG.
After searching for an option to run Java code from Django application(python), I found out that Py4J is the best option for me. I tried Jython, JPype and Python subprocess and each of them have certain limitations:
Jython. My app runs in python.
JPype is buggy. You can start JVM just once after that it fails to start again.
Python subprocess. Cannot pass Java object between Python and Java, because of regular console call.
On Py4J web site is written:
In terms of performance, Py4J has a bigger overhead than both of the previous solutions (Jython and JPype) because it relies on sockets, but if performance is critical to your application, accessing Java objects from Python programs might not be the best idea.
In my application performance is critical, because I'm working with Machine learning framework Mahout. My question is: Will Mahout also run slower because of Py4J gateway server or this overhead just mean that invoking Java methods from Python functions is slower (in latter case performance of Mahout will not be a problem and I can use Py4J).
JPype issue that #HIP_HOP mentioned with JVM getting detached from new threads can be overcome with the following hack (add it before the first call to Java objects in the new thread which does not have JVM yet):
# ensure that current thread is attached to JVM
# (essential to prevent JVM / entire container crashes
# due to "JPJavaEnv::FindClass" errors)
if not jpype.isThreadAttachedToJVM():
jpype.attachThreadToJVM()
PySpark uses Py4J quite successfully. If all the heavylifting is done on Spark (or Mahout in your case) itself, and you just want to return result back to "driver"/Python code, then Py4J might work for you very well as well.
Py4j has slightly bigger overhead for huge results (that's not necessarily the case for Spark workloads, as you only return summaries /aggregates for the dataframes). There is an improvement discussion for py4j to switch to binary serialization to remove that overhead for higher badnwidth requirements too: https://github.com/bartdag/py4j/issues/159
My solutions
java thread/process <-> Pipes <-> py subprocess
Use pipes by java's ProcessBuilder to call py with args "-u" to transfer data via pipes.
Here is a good practice.
https://github.com/JULIELab/java-stdio-ipc
Here is my stupid research result about "java <-> py"
[Jython] Java implement of python.
[Jpype] JPype is designed to allow the user to exercise Java as fluidly as possible from within Python. We can break this down into a few specific design goals.
Unlike Jython, JPype does not achieve this by re-implementing Python, but instead by interfacing both virtual machines at the native level. This shared memory based approach achieves good computing performance while providing the access to the entirety of CPython and Java libraries.
[Runtime] The Runtime class in java (old method).
[Process] Java ProcessBuilder class gives more structure to the arguments.
[Pipes] Named pipes could be the answer for you. Use subprocess. Popen to start the Java process and establish pipes to communicate with it.
Try mkfifo() implementation in python.
https://jj09.net/interprocess-communication-python-java/
-> java<-> Pipes <-> py https://github.com/JULIELab/java-stdio-ipc
[Protobuf] This is the opensource solution Google uses to do IPC between Java and Python. For serializing and deserializing data efficiently in a language-neutral, platform-neutral, extensible way, take a look at Protocol Buffers.
[Socket] CS-arch throgh socket
Server(Python) - Client(Java) communication using sockets
https://jj09.net/interprocess-communication-python-java/
Send File From Python Server to Java Client
[procbridge]
A super-lightweight IPC (Inter-Process Communication) protocol over TCP socket. https://github.com/gongzhang/procbridge
https://github.com/gongzhang/procbridge-python
https://github.com/gongzhang/procbridge-java
[hessian binary web service protocol] using python client and java server.
[Jython]
Jython is a reimplementation of Python in Java. As a result it has much lower costs to share data structures between Java and Python and potentially much higher level of integration. Noted downsides of Jython are that it has lagged well behind the state of the art in Python; it has a limited selection of modules that can be used; and the Python object thrashing is not particularly well fit in Java virtual machine leading to some known performance issues.
[Py4J]
Py4J uses a remote tunnel to operate the JVM. This has the advantage that the remote JVM does not share the same memory space and multiple JVMs can be controlled. It provides a fairly general API, but the overall integration to Python is as one would expect when operating a remote channel operating more like an RPC front-end. It seems well documented and capable. Although I haven’t done benchmarking, a remote access JVM will have a transfer penalty when moving data.
[Jep]
Jep stands for Java embedded Python. It is a mirror image of JPype. Rather that focusing on accessing Java from within Python, this project is geared towards allowing Java to access Python as a sub-interpreter. The syntax for accessing Java resources from within the embedded Python is quite similar to support for imports. Notable downsides are that although Python supports multiple interpreters many Python modules do not, thus some of the advantages of the use of Python may be hard to realize. In addition, the documentation is a bit underwhelming thus it is difficult to see how capable it is from the limited examples.
[PyJnius]
PyJnius is another Python to Java only bridge. Syntax is somewhat similar to JPype in that classes can be loaded in and then have mostly Java native syntax. Like JPype, it provides an ability to customize Java classes so that they appear more like native classes. PyJnius seems to be focused on Android. It is written using Cython .pxi files for speed. It does not include a method to represent primitive arrays, thus Python list must be converted whenever an array needs to be passed as an argument or a return. This seems pretty prohibitive for scientific code. PyJnius appears is still in active development.
[Javabridge]
Javabridge is direct low level JNI control from Python. The integration level is quite low on this, but it does serve the purpose of providing the JNI API to Python rather than attempting to wrap Java in a Python skin. The downside being of course you would really have to know a lot of JNI to make effective use of it.
[jpy]
This is the most similar package to JPype in terms of project goals. They have achieved more capabilities in terms of a Java from Python than JPype which does not support any reverse capabilities. It is currently unclear if this project is still active as the most recent release is dated 2014. The integration level with Python is fairly low currently though what they do provide is a similar API to JPype.
[JCC]
JCC is a C++ code generator that produces a C++ object interface wrapping a Java library via Java’s Native Interface (JNI). JCC also generates C++ wrappers that conform to Python’s C type system making the instances of Java classes directly available to a Python interpreter. This may be handy if your goal is not to make use of all of Java but rather have a specific library exposed to Python.
[VOC] https://beeware.org/project/projects/bridges/voc/_
A transpiler that converts Python bytecode into Java bytecode part of the BeeWare project. This may be useful if getting a smallish piece of Python code hooked into Java. It currently list itself as early development. This is more in the reverse direction as its goals are making Python code available in Java rather providing interaction between the two.
[p2j]
This lists itself as “A (restricted) python to java source translator”. Appears to try to convert Python code into Java. Has not been actively maintained since 2013. Like VOC this is primilarly for code translation rather that bridging.
[GraalVM]
Source: https://github.com/oracle/graal
I don't know Mahout. But think about that: At least with JPype and Py4J you will have performance impact when converting types from Java to Python and vice versa. Try to minimize calls between the languages. Maybe it's an alternative for you to code a thin wrapper in Java that condenses many Javacalls to one python2java call.
Because the performance is also a question about your usage screnario (how often you call the script and how large is the data that is moved) and because the different solutions have their own specific benefits/drawbacks, I have created an API to switch between different implementations without you having to change your python script: https://github.com/subes/invesdwin-context-python
Thus testing what works best or just being flexible about what to deploy to is really easy.
I have a quite huge Java class that has several imported packages and libraries (related to natural language processing). I want to call some specific methods of my Java classes and get results back using a Perl script. How should I do this? I guess the Inline::Java is not suitable for my purpose since the Java code is quite large to be incorporated at one place.
Why is Inline::Java not suitable? You don't need to provide access to all the bits of your Java API. You can write short methods that call into your Java stuff. A small Java adapter layer can show up in your Perl code to give you the access you need. Have you tried it yet?
How big is this Java code base? I've been on projects that easily integrated big Java SDKs (although I wasn't that one doing that part).
I had ran into similar situation lately. The best solution which worked for me was to use apache thrift service and expose the required methods through it. These method were then in turn consumed by the client written in perl.