how to isolate JNA bindings that are platform specific? - java

I have a project where I defined a JNA wrapper to Windows kernel32 library, upon which I have made several helpers that are not critical for the project but increase the integration to the platform (namely: system debug logging with OutputDebugString + DebugView and Mailslot messaging features).
Here is my JNA defines:
import java.nio.ByteBuffer;
import java.util.HashMap;
import java.util.Map;
import com.sun.jna.Native;
import com.sun.jna.ptr.IntByReference;
import com.sun.jna.win32.StdCallLibrary;
import com.sun.jna.win32.W32APIFunctionMapper;
import com.sun.jna.win32.W32APITypeMapper;
public interface JnaKernel32 extends StdCallLibrary {
//StdCall is needed for lib kernel32
#SuppressWarnings("unchecked")
Map ASCII_OPTIONS = new HashMap(){
{
put(OPTION_TYPE_MAPPER, W32APITypeMapper.ASCII);
put(OPTION_FUNCTION_MAPPER, W32APIFunctionMapper.ASCII);
}
};
#SuppressWarnings("unchecked")
Map UNICODE_OPTIONS = new HashMap(){
{
put(OPTION_TYPE_MAPPER, W32APITypeMapper.UNICODE);
put(OPTION_FUNCTION_MAPPER, W32APIFunctionMapper.UNICODE);
}
};
Map DEFAULT_OPTIONS = Boolean.getBoolean("w32.ascii") ? ASCII_OPTIONS : UNICODE_OPTIONS;
JnaKernel32 INSTANCE = (JnaKernel32) Native.loadLibrary("kernel32", JnaKernel32.class, DEFAULT_OPTIONS);
//some system defines
//...
}
And the Mailslot definition:
public class Mailslot {
static JnaKernel32 kernel32 = JnaKernel32.INSTANCE;
boolean localMailslot = false;
int lastError = 0;
private int hMailslot = JnaKernel32.INVALID_HANDLE_VALUE;
//...
}
And in some places I have also
static JnaKernel32 kernel32 = JnaKernel32.INSTANCE; //to call OutputDebugString
//...
kernel32.OutputDebugString("some debug message");
My concern is that project could be used also on GNU/Linux or MacOS X but obviously the Native.loadLibrary fails at runtime if executed on e.g. OSX.
I am considering about
either porting the native feature with other JNA binding
or simply disable the existing Windows kernel32 binding when running on another platform as it is just convenient but not mandatory helpers.
How can I isolate the platform-specific features and calls that are made? I was thinking about moving the JNA part to a runtime loaded plugin perhaps?

The answer is really a general software development strategy.
Identify the API that your client code needs
In this case, it might be MailsSlot.sendMessage(int destID, String msg)
Abstract the implementation details behind that API
public interface MailSlot {
void sendMessage(int destId, String msg);
}
Provide one or more concrete implementations to meet the API contract
public class Win32MailSlot implements MailSlot {
public void sendMessage(int destId, String msg) {
// Do stuff here that's windows-specific
}
}
public class OSXMailSlot implements MailSlot {
public void sendMessage(int destId, String msg) {
// Do stuff here that's windows-specific
}
}
Choose the appropriate implementation at runtime:
MailSlot mslot = Platform.IS_WINDOWS ? new Win32MailSlot() : new OSXMailSlot();
After a few implementations, you may find some duplicated code, which you might then refactor into an abstract base class shared among the platform-specific implementations.
See the JNA platform FileUtils class for an example of such an strategy.

Related

org.openide.util.Lookup Cannot Find Any Classes Implementing

SQLUtils.java:
import org.openide.util.Lookup;
import java.util.ServiceLoader; // This doesn't work either
public class SQLUtils {
public static DBDriver getDriver(String prefix) {
for(DBDriver e : Lookup.getDefault().lookupAll(DBDriver.class)) {
System.out.println(e.getPrefix());
if(e.getPrefix().equalsIgnoreCase(prefix)) {
return e;
}
}
return null;
}
}
MySQLDriver.java:
public class MySQLDriver implements DBDriver {
#Override
public String getPrefix() {
return "mysql";
}
}
DBDriver.java:
import java.io.Serializable;
public interface DBDriver extends Serializable {
public String getPrefix();
}
Main.java:
public class Main {
public static void main(String[] args) {
DBDriver d = SQLUtils.getDriver("mysql");
}
}
This does nothing when running it, it cannot find any classes implementing.
What the program is trying to do is get the driver that is entered as a parameter for SQLUtils.getDriver(String prefix) (in Main.java).
For some reason I cannot get this to work.
I'm not familiar with OpenIDE Lookup mechanism, but I am familiar with the Java ServiceLoader mechanism.
You need to provide a file in the META-INF/services/ folder describing what classes implement specific interfaces. From the Java Docs describing the ServiceLoader class is this example:
If com.example.impl.StandardCodecs is an implementation of the
com.example.CodecSet service then its jar file also contains a file
named
META-INF/services/com.example.CodecSet
This file contains the single line:
com.example.impl.StandardCodecs # Standard codecs implementing com.example.CodecSet
What you are missing is a similar file that needs to be included on your classpath or within your JAR file.
You don't include you package names so I cannot provide a more direct example to help solve your problem.
I dropped the NetBeans API and switched to Reflections. I implemented Maven and ran it with IntelliJ. Works well for me.

Dynamic Typecasting in Java

I'm writing a plugin for the Minecraft server implementation CraftBukkit, and I've come across a problem where I need to cast to a class that is found through reflection.
Here's the deal. The original code I wrote looked like this, with irrelevant parts removed:
import java.util.ArrayList;
import java.util.Collection;
import java.util.Iterator;
import java.util.List;
import java.util.Random;
import net.minecraft.server.v1_7_R3.EntityAnimal;
import net.minecraft.server.v1_7_R3.EntityHuman;
import org.bukkit.craftbukkit.v1_7_R3.entity.CraftAnimals;
import org.bukkit.craftbukkit.v1_7_R3.entity.CrafteEntity;
import org.bukkit.World;
import org.bukkit.entity.Animals;
import org.bukkit.entity.Entity;
import org.bukkit.event.Listener;
import org.bukkit.plugin.java.JavaPlugin;
import org.bukkit.scheduler.BukkitRunnable;
public class Task extends BukkitRunnable {
private static final int MATING_DISTANCE = 14;
private final JavaPlugin plugin;
private final Random randomizer;
private boolean mateMode;
private double chance;
public Task(JavaPlugin plugin, double chance, boolean mateMode) {
this.plugin = plugin;
this.randomizer = new Random();
this.chance = chance;
this.mateMode = mateMode;
this.theTaskListener = listener;
}
public void run() {
List<World> worlds = plugin.getServer().getWorlds();
Iterator<World> worldIterator = worlds.iterator();
while (worldIterator.hasNext()) {
World world = worldIterator.next();
Collection<Animals> animals = world.getEntitiesByClass(Animals.class);
Iterator<Animals> animalIterator = animals.iterator();
while (animalIterator.hasNext()) {
Animals animal = (Animals) animalIterator.next();
EntityAnimal entity = (EntityAnimal) ((CraftEntity) ((CraftAnimals) animal)).getHandle();
EntityHuman feeder = null;
entity.f(feeder);
}
}
}
}
However, as you can see in the imports, this code imported the classes from only one version of the Minecraft server package - v1_7_R3. Now the problem is, I want to add support for more than that, and I want to be able to do that without creating separate versions of my plugin for each version of Minecraft. Despite the fact that most of the classes in the package are the same (at least ALL of those that I need) the package names are different, and therefore it can't be done with static imports (or at least I think so?)
So, I decided to use reflection in order to get the correct classes I need (this code is in another class):
private static final String[] requiredClasses = {
"net.minecraft.server.%s.EntityAnimal",
"net.minecraft.server.%s.EntityHuman",
"org.bukkit.craftbukkit.%s.entity.CraftAnimals",
"org.bukkit.craftbukkit.%s.entity.CraftEntity"
};
public static final String[] supportedVersions = {
"v1_7_R3",
"v1_7_R4"
};
public Class<?>[] initializeClasses() {
String correctVersion = null;
for (int i = 0; i < supportedVersions.length; i++) {
String version = supportedVersions[i];
boolean hadIssues = false;
for (int j = 0; j < requiredClasses.length; j++) {
String className = requiredClasses[j];
try {
Class.forName(String.format(className, version));
} catch (ClassNotFoundException e) {
getLogger().log(Level.INFO, String.format("The correct version isn't %s.", version));
hadIssues = true;
break;
}
}
if (!hadIssues) {
correctVersion = version;
break;
}
}
Class[] classes = new Class[requiredClasses.length];
if (correctVersion != null) {
getLogger().log(Level.INFO, String.format("The correct version is %s.", correctVersion));
for (int i = 0; i < requiredClasses.length; i++) {
String className = requiredClasses[i];
try {
classes[i] = Class.forName(String.format(className, correctVersion));
} catch (ClassNotFoundException e) {}
}
} else {
getLogger().log(Level.WARNING, "The version of Minecraft on this server is not supported.");
getLogger().log(Level.WARNING, "Due to this, the plugin will self-disable.");
getLogger().log(Level.WARNING, "To fix this issue, get build that supports your version.");
this.setEnabled(false);
}
return classes;
}
Now, this approach successfully retrieves the required classes in both versions currently supported. I passed these to the rewritten Task class using instance variables and an edited constructor, and I removed the version-specific imports:
import java.util.ArrayList;
import java.util.Collection;
import java.util.Iterator;
import java.util.List;
import java.util.Random;
import org.bukkit.World;
import org.bukkit.entity.Animals;
import org.bukkit.entity.Entity;
import org.bukkit.plugin.java.JavaPlugin;
import org.bukkit.scheduler.BukkitRunnable;
public class Task extends BukkitRunnable {
private static final int MATING_DISTANCE = 14;
private final JavaPlugin plugin;
private final Random randomizer;
private boolean mateMode;
private double chance;
private Class entityAnimal;
private Class entityHuman;
private Class craftAnimals;
private Class craftEntity;
public Task(JavaPlugin plugin, Class[] classes, double chance, boolean mateMode) {
this.plugin = plugin;
this.randomizer = new Random();
this.chance = chance;
this.mateMode = mateMode;
this.entityAnimal = classes[0];
this.entityHuman = classes[1];
this.craftAnimals = classes[2];
this.craftEntity = classes[3];
}
Now, how can I rewrite the Task.run() method so that it will use the reflection classes? There is a whole lot of typecasting involved and unfortunately it's all necessary due to the ridiculous amount of overloading in the Minecraft code. For example, the entity.f(EntityHuman human) method cannot simply be called by doing entity.f(null) because there are other overloading entity.f(Object object) methods.
I am open to all suggestions as I'm facing a dead-end here. If there is a better approach to the problem, I could change to that as well.
Thank you!
In an object oriented language, we have access to various design patterns that have been developed for exactly this purpose. We use two patterns in particular.
Adapter Pattern is used to provide the same interface to a number of different implementations. It is sometimes called a shim. You create one class per version of each server, importing libraries to each. The class implements an interface that they hold in common.
Factory Pattern is used to select among the adapter classes. You use whatever method you need to determine which server version you have, and it will create an object implementing the proper interface. The main code remains the same. It calls the factory to get an object that knows how to deal with the server.
The advantages of this approach are several. You don't pollute the name space by importing overlapping libraries. The main code is much less susceptible to change as new server versions are added; the only code that needs to be written is the new server shim and the factory that determines which adapter to produce.
Just a brainstorming idea. What if:
importing all supported versions
fully referencing the appropriate package's types
checking for the version that's targeted at a particular runtime (assumed it can be obtained somehow)
import net.minecraft.server.v1_7_R3.*;
import net.minecraft.server.v1_7_R4.*;
enum Version {
V1_7_R3,
V1_7_R4
}
Version currentVersion;
net.minecraft.server.v1_7_R3.EntityAnimal animal3;
net.minecraft.server.v1_7_R4.EntityAnimal animal4;
// obtain currentVersion
switch ( currentVersion ) {
case V1_7_R3:
animal3.method();
break;
case V1_7_R4:
animal4.method();
break;
}
This is somehow ugly, of course, but under the given circumstances it's the possibility that came into my mind first.
After reading Gerold Broser's response, I realized that I would have to somehow modify my approach in order to create some sort of a handler class that would carry out the version-specific operation - of course this would be an interface that would separately be implemented by a class per version.
However, this became a problem when I realized Maven wouldn't let me call two versions of the same groupid.artifactid object.
I quickly did some research and found mbaxter's Multiple Versions Tutorial as well as the AbstractionExamplePlugin implementation which perfectly demonstrates this approach.
The approach works perfectly and is what every Bukkit developer should use. Here's my finished plugin for further reference if necessary.

Java JNA Different implementation according to OS

I have to use Java JNA to link a C library. This library has a Windows implementation and a Linux one. These differ one from other for a single method because this method is implemented only by Windows version.
MyJnaInterface INSTANCE = (MyJnaInterface) Native.loadLibrary("MyLibrary",
MyJnaInterface.class);
I would like to have just one version of my Java application, this may have a single interface with 2 implementation, one for windows os and one for linux os, obviously the linux implementation will have an empty method.
public interface MyJnaInterface
public class MyJnaWinImpl implements MyJnaInterface
public class MyJnaLinuxImpl implements MyJnaInterface
This works in windows, in linux OS at service startup JNA tries to find its native methods also in windows classes (also if this class is not used in runtime) and so it throws a UnsatifiedLinkError.
How to solve this deadlock?
I really cannot change the native library (it would be so simple...)
I suggest to use compilation toolbox in your project to compile java code runtime depending upon value returned by System.getProperty("os.name") . If it returns windows then you can add source code for MyJnaWinImpl in one string and pass that to JavaSourceCompiler class. Once that is compiled load class and create instance. On linux JavaSourceCompiler will compile MyJnaLinuxImpl. Ensure that before creating this instance libraries are loaded.
Below is a small test code snippet.
package test;
import org.abstractmeta.toolbox.compilation.compiler.*;
import org.abstractmeta.toolbox.compilation.compiler.impl.*;
import java.lang.ClassLoader;;
public class test {
public static void main(String[] args) throws ClassNotFoundException,InstantiationException,IllegalAccessException{
JavaSourceCompiler javaSourceCompiler = new JavaSourceCompilerImpl();
JavaSourceCompiler.CompilationUnit compilationUnit = javaSourceCompiler.createCompilationUnit();
String os = System.getProperty("os.name");
String SourceCode;
if ( os.contentEquals("Windows"))
{
SourceCode = "package com.test.foo;\n" +
"import MyJnaInterface.*;" +
"import MyJnaWinImpl " +
"public class Foo implements MyJnaWinImpl {\n" +
" public native void check();\n" +
" }";
}
else
{
SourceCode = "package com.test.foo;\n" +
"import MyJnaInterface.*;" +
"import MyJnaLinuxImpl " +
"public class Foo implements MyJnaLinuxImpl {\n" +
//" public native void check();\n" +
" }";
}
compilationUnit.addJavaSource("com.test.foo.Foo", SourceCode);
ClassLoader classLoader = javaSourceCompiler.compile(compilationUnit);
Class fooClass = classLoader.loadClass("com.test.foo.Foo");
Object foo = fooClass.newInstance();
}
}
I solved using static{} block.
public interface MyJnaInterface;
public interface MyJnaInterfaceWin implements MyJnaInterface; // this has the WinMethod method
...
private static MyJnaInterface INSTANCE;
static{
if(SystemUtils.IS_OS_LINUX){
INSTANCE=(MyJnaInterface) Native.loadLibrary("MyLibrary",MyJnaInterface.class);
}else{
INSTANCE=(MyJnaInterfaceWin) Native.loadLibrary("MyLibrary",MyJnaInterfaceWin.class);
}
}
...
public static void WinMethod(){
if(!SystemUtils.IS_OS_LINUX) ((MyJnaInterfaceWin)INSTANCE).WinMethod());
}
I'm assuming you're using direct mapping, since interface mapping won't look up your function until you invoke it.
Write a base class with the base implementation, then a derived class that includes the additional mapping. Only load the derived class where you know the underlying function exists.
class BaseInterface {
public native void nativeMethod();
public void extendedMethod() { /* empty stub */ }
}
class ExtendedInterface extends BaseInterface {
public native void extendedMethod();
}
if (needExtendedInterface) {
lib = /* load extended library */
}
else {
lib = /* load base library */
}

Call DLL from Java using JNA

I am new to accessing DLLs from Java using JNA. I need to access methods from a class within a DLL(written in .net). Form this sample DLL below, I am trying to get AuditID and Server ID. I am ending with the following error while I am running my code. Any guidance really appreciated.
/// Error ///
Exception in thread "main" java.lang.UnsatisfiedLinkError: Error looking up function 'GetEnrollcontext': The specified procedure could not be found.
//DLL File Code//
SampleDLL.ProfileEnroll enrollcontext = new SampleDLL.ProfileEnroll();
enrollcontext.Url =” url”;
enrollcontext.AuditIdType = SampleDLL.ProfileId;
enrollcontext.AuditId = “22222222 “;
enrollcontext.ServerId = “server1”;
/// Java Code ///
import com.sun.jna.Library;
import com.sun.jna.Native;
import com.sun.jna.Structure;
import dllExtract.DLLExtractTest.SampleDLL.Enrollcontext;
public class SampleDLLExtract {
public interface SampleDLL extends Library {
SampleDLL INSTANCE = (SampleDLL) Native.loadLibrary("SampleDLL",
SampleDLL.class);
public static class Enrollcontext extends Structure {
public String auditId;
public String serverId;
}
void GetEnrollcontext(Enrollcontext ec); // void ();
}
public static void main(String[] args) {
SampleDLL sdll = SampleDLL.INSTANCE;
SampleDLL.Enrollcontext enrollContext = new SampleDLL.Enrollcontext();
sdll.GetEnrollcontext(enrollContext);
System.out.println(sdll.toString(sdll.GetEnrollcontext(enrollContext)));
}
}
in fact there is a solution for you to use C#, VB.NET or F# code via JNA in Java (and nothing else)! and it is also very easy to use:
https://www.nuget.org/packages/UnmanagedExports
with this package all you need to do is, add [RGiesecke.DllExport.DllExport] to your methods like that:
C# .dll Project:
[RGiesecke.DllExport.DllExport]
public static String yourFunction(String yourParameter)
{
return "CSharp String";
}
Java Project:
public interface jna extends Library {
jna INSTANCE = (jna) Native.loadLibrary("yourCSharpProject.dll", jna.class);
public String yourFunction(String yourParameter);
}
use it in the code:
System.out.println(jna.INSTANCE.yourFunction("nothingImportant"));
Viola!
As already mentioned it works very easy, but this solution has some limitations:
only available for simple datatypes as parameter & return values
no MethodOverloading available. yourFunction(String yourParameter) and yourFunction(String yourParameter, String yourSecondParameter) does not work! you have to name them differently
Use arrays as parameter or return values. (JNA offers StringArray, but I am not able to use them in C#) (maybe there is a solution, but I couldn't come up with one so far!)
if you export a method you can't call it internally in your C# code (simple to bypass that by the following:
.
[RGiesecke.DllExport.DllExport]
public static Boolean externalAvailable(String yourParameter)
{
return yourInternalFunction(yourParameter);
}
With C# it works great, with VB.NET and F# I have no experience.
hope this helps!

How to get the Groovy generated java source code

We have some legacy code with Groovy, and we want to remove Groovy from the application, so, we need to get the java source code generated after using the gmaven plug-in.
Basically, in other words I am dynamically generating new classes (using gmaven Groovy maven plug in) and I would like to be able to obtain the java source code of such generated classes.
I researched a little bit and can see that the only goals for this plug in are
<goal>generateStubs</goal>
<goal>compile</goal>
<goal>generateTestStubs</goal>
<goal>testCompile</goal>
I can't see any goal that allows you to obtain the fully implemented java source code, the stub code is not enough for us as we need the final implementation source code in order to get rid of Groovy.
I'm not very familiar with the gmaven plugin, but I assume it compiles the groovy code into byte code. In this case, you can use a byte code decompiler, there is a nice list here. In the past I've used JAD and it was quite nice. The best ones will also try to create meaningful variable names based on class names.
One warning though - Groovy objects are derived from GObject, not java.lang.Object, so you would probably need to keep the groovy jar until the groovy->java porting is done. Also, be prepared that it won't be a very easy to read java...
It may be out of your scope (1 year old) but I fought against the same problem and found a method to retrieve the algorithm (not the java source code) from the decompiled groovy classes.
You may want to take a look : http://michael.laffargue.fr/blog/2013/11/02/decompiling-groovy-made-classes/
The generated stubs will be useless for you. They are just what their names suggests: stubs.
The stubs are only useful when doing joint java/groovy compilation. That's because there are two compilers involved in a java/groovy mixed project.
Parse groovy
Create stubs
Compile java and stubs (using javac)
Continue groovy compilation (using groovyc)
The groovy code will be compiled using groovyc compiler and the result is byte code.
This is an example of a generated stub:
package maba.groovy;
import java.lang.*;
import java.io.*;
import java.net.*;
import java.util.*;
import groovy.lang.*;
import groovy.util.*;
#groovy.util.logging.Log4j() public class Order
extends java.lang.Object implements
groovy.lang.GroovyObject {
public groovy.lang.MetaClass getMetaClass() { return (groovy.lang.MetaClass)null;}
public void setMetaClass(groovy.lang.MetaClass mc) { }
public java.lang.Object invokeMethod(java.lang.String method, java.lang.Object arguments) { return null;}
public java.lang.Object getProperty(java.lang.String property) { return null;}
public void setProperty(java.lang.String property, java.lang.Object value) { }
public int getPrice() { return (int)0;}
public void setPrice(int value) { }
public int getQuantity() { return (int)0;}
public void setQuantity(int value) { }
#java.lang.Override() public java.lang.String toString() { return (java.lang.String)null;}
}
As you can see there is nothing useful. And you will still depend on some groovy libraries.
This question has been on the mailing-list some time ago [0]. To summarize: Groovy to Java is hard to achieve since there are language constructs and APIs (if you do want to totally remove the Groovy dependency) that are not available in Java.
Especially with the introduction of call-site caching and other performance optimizing techniques the generated Java code would look a lot like this (for the matter of simplicity I just threw some script into JD-GUI [1]):
public class script1351632333660 extends Script
{
public script1351632333660()
{
script1351632333660 this;
CallSite[] arrayOfCallSite = $getCallSiteArray();
}
public script1351632333660(Binding arg1)
{
Binding context;
CallSite[] arrayOfCallSite = $getCallSiteArray();
ScriptBytecodeAdapter.invokeMethodOnSuperN($get$$class$groovy$lang$Script(), this, "setBinding", new Object[] { context });
}
public Object run()
{
CallSite[] arrayOfCallSite = $getCallSiteArray(); Object items = ScriptBytecodeAdapter.createList(new Object[0]);
Object[] item = (Object[])ScriptBytecodeAdapter.castToType(ScriptBytecodeAdapter.createList(new Object[] { "Fluff", arrayOfCallSite[1].callConstructor($get$$class$java$util$Date()), (Integer)DefaultTypeTransformation.box(11235813) }), $get$array$$class$java$lang$Object());
arrayOfCallSite[2].call(items, item);
arrayOfCallSite[3].callCurrent(this, items);
ValueRecorder localValueRecorder = new ValueRecorder();
try
{
Object tmp102_101 = items; localValueRecorder.record(tmp102_101, 8);
Object tmp126_121 = arrayOfCallSite[4].call(tmp102_101, new script1351632333660._run_closure1(this)); localValueRecorder.record(tmp126_121, 14); if (DefaultTypeTransformation.booleanUnbox(tmp126_121)) localValueRecorder.clear(); else ScriptBytecodeAdapter.assertFailed(AssertionRenderer.render("assert items.findAll { it }", localValueRecorder), null); } finally {
localValueRecorder.clear(); throw finally; } return null; return null; }
static { __$swapInit();
Long localLong1 = (Long)DefaultTypeTransformation.box(0L);
__timeStamp__239_neverHappen1351632333665 = localLong1.longValue();
Long localLong2 = (Long)DefaultTypeTransformation.box(1351632333665L);
__timeStamp = localLong2.longValue(); }
class _run_closure1 extends Closure implements GeneratedClosure { public _run_closure1(Object _thisObject) { super(_thisObject); }
public Object doCall(Object it) { CallSite[] arrayOfCallSite = $getCallSiteArray(); return it; return null;
}
// ...
[0] http://groovy.329449.n5.nabble.com/Java-lt-gt-Groovy-converters-td337442.html
[1] http://java.decompiler.free.fr

Categories