Welcome!

Java Authors: Elizabeth White, Pat Romanski, Liz McMillan, Jason Bloomberg, Trevor Parsons

Related Topics: Java

Java: Article

Embedding the Java Virtual Machine, Once And For All

A solution with complete support

JTL is a portable C++ JNI template library that allows you to embed a JVM within another program. An example of this might be found inside a browser that needs to support Java plug-ins. First I'll discuss the traditional way to embed a JVM within a native program, and then talk about the generic solution offered by JTL.

To apply the solution provided by JTL, you need to know the basics of Java, C++, and JNI. However, to understand JTL's design you should be familiar with modern C++ techniques, such as template programming and the Boost library (www.boost.org).

All the sample code in this article is written for Microsoft Windows, but it can easily be ported to other platforms.

The Traditional Way to Embed the JVM

Although there are lots of resources on how to embed a JVM, most of them are based on Sheng Liang's classic book, Java Native Interface: Programmer's Guide and Specification. Figure 1 shows the basic procedure.

If the application has other threads that are using the JVM as well, the communications among these threads will look like Figure 2.

Listing 1 shows the typical code (all the error checking and exception handling code in this article has been omitted for clarity). This code works fine as long as your program is simple enough; however, there are three major drawbacks.

Issue 1: The JVM is a global variable, which is not acceptable in some cases.

Issue 2: The threads in which jvm and env are used and the thread that launches jvm couple tightly. For example, there will be a potential timing issue that's clearly shown in Figure 2. The AttachCurrentThread(...) must be called after the launcher thread initialized jvm, and DetachCurrentThread() must be called before jvm has been destroyed; otherwise this creates unnecessary communications overhead between threads. It gets worse when you don't know which thread is the launcher thread.

Of course the VM can always be created in one thread when the native application starts. However, this is a pure waste if there is no Java client. The JVM should not be created if there are no requests.

Issue 3. The function calls to Attach-CurrentThread(...) and Detach-Current Thread() should be paired so the JVM can get a chance to free the local references. However, it's too easy to not have the DetachCurrentThread() called by the programmer (because of multiple flow paths) or by exceptions. A better approach would be to encapsulate the JVM into a class or classes.

Thin JVM Wrapper

It's natural to wrap the JavaVM* pointer into a member variable, say jvm. Because only one JVM in each process can be created (until JDK 1.4.2), the obvious class design is to make jvm static. Listing 2 shows the first attempt.

Here the method's signatures are not important so let's focus on the class design. The thin wrapper approach is simple and straightforward; however, it's not much better than the C-like code in Listing 1. It only gets rid of the explicit global variable jvm (the static member variable is still kind of global though). Issues 2 and 3, mentioned in the previous section, still need to be addressed.

Singleton JVM

The Singleton design pattern is described in Design Patterns by Erich Gamma, et al, as "Ensure a class only has one instance, and provide a global point of access of it." In other words, a Singleton class is a class for which no more than one instance can exist at runtime. This is what the JVM behaves like. We can apply the Singleton pattern to the JVM class design.

Andrei Alexandrescu has provided all kinds of singleton implementations in his book Modern C++ Design and the Loki library. Loki has a singleton manager class template, SingletonHolder, that looks like Listing 3.

The template parameter T in Listing 3 is the target class, in our case CJavaVM. Other template parameters specify the policies that manage the singleton. As explained in Alexandrescu's book, "A policy defines a class interface or a class template interface. The interface consists of one or all of the following: inner type definitions, member functions, and member variables." Policy classes are not intended for stand-alone use and their member functions are often static. In Loki::-Single-tonHolder, the CreationPolicy determines how the singleton instance is created and destroyed, and the Threading-Model policy determines whether the singleton is living in a multiple threaded world.

Listing 4 provides the redesigned CJavaVM class and its usage with Loki::-SingletonHolder. A difference between CJavaVM and CJavaVM2 is that all members in CJavaVM2 are not static. Listing 4 looks fancier than Listing 1. However, it does not solve the real problems. Issues 2 and 3 are still unresolved. For example, if there are any other threads trying to make a JNI call, it's still a requirement that the JVM launcher thread has already called jvm.Start-JavaVM(). This is because CJavaVM2's constructor does nothing. If StartJava-VM() can be moved into CJavaVM2's constructor and Destroy-JavaVM() can be moved into CJavaVM2's destructor, then the timing issue will be solved, which is not easy to do.

The major difficulty is that Loki::-SingletonHolder's template parameter CreationPolicy only calls the target class T's default constructor. In other words, CJavaVM2's constructor cannot look something like this: CJavaVM2(std::-string- jvmpath, JavaVMInitArgs& args). It must be CJavaVM2(). Period.

We are not out of bullets though. The "Every problem can be solved by adding another layer of indirection" idiom applies here.

Policy-Based JVM Class

The solution is to make CJavaVM2 policy-based too. Listing 5 shows the new version.

In CJavaVM3, a new policy class JavaVMLauncher is introduced and declared as follows:


class JavaVMLauncher {
public:
    static void LaunchJavaVM(void** ppLib, JavaVM** ppJavaVM);
    static void DestroyJavaVM(void* pLib, JavaVM* pJavaVM);
}

ppLib is a pointer to a handle returned by LoadLibrary on Windows or dlopen on Solaris. To make JavaVMLauncher more generic, the void pointer is used here. The pointer to pointer method is used for ppLib and ppJavaVM in LaunchJavaVM() because JavaVMLauncher does not contain any member variables. It's the caller to provide the placeholders for the objects pointed to by ppLib and ppJavaVM. DestroyJavaVM() simply accepts the cached JavaVM* and lib handle to destroy the JVM and free the library.

Now we can invoke the JVM as in Listing 6, which is pretty much what JTL does. In JTL, there is another class template called JavaVMMgr that is the placeholder for the JavaVM* and pLib. All the JVM calls will be delegated to the JavaVMMgr.

Exception Handling

The LaunchJavaVM() method in JavaVMLauncher can fail; for example, if the user does not have a JRE installed on his machine. When this happens, LaunchJavaVM() can return a flag or error code. In JTL, it will simply throw a jtl::load_javavm_error exception.

Multithreading, Extending Loki, and boost::thread

Remember in Listing 3, Loki::Single-tonHolder has a template parameter ThreadingModel that's used to synchronize the calls to MakeInstance(). However, under different platforms the thread implementation and API are different. To make the JVM class design platform portable, it needs to use a portable threading framework. JTL uses boost::thread because it's simple and easy to use, and it will likely be part of the next version of the standard C++ library.

Though Alexandrescu provides us with the excellent SingletonHolder class, JTL does not use it directly for two main reasons: not all compilers support Loki well and Loki::ThreadingModel is only implemented to work with Windows.jtl::SingletonHolder does some extension to Loki::SingletonHolder to achieve the platform-portable goal. However, jtl::SingletonHolder uses the same template parameters (and in the same order) as Loki::SingletonHolder, therefore the user can easily switch to Loki's version if someday Loki provides us with a platform-portable ThreadingModel.

Another thing needs to be mentioned for ThreadingModel - it should be able to work with both the single-thread model and multiple-thread model. If there's only one thread, there's no sense in doing the synchronization. To achieve this generic solution, JTL provides two class templates, Multiple-ThreadModel and SingleThread-Model, as shown in Listing 7.

These two classes simply provide type definitions. In Mul-tipleThread-Model, JTL uses the boost::mutex and boost::mutex::scoped_lock as the synchronization primi-tives, while in Single-ThreadModel it uses boostex::-faked_-mutex and boostex::faked_mutex::-scoped_-lock as the synchronization primitives. The faked mutex and scoped_lock are simple extensions to boost mutex and scoped_lock. They're empty classes and provide only the required inter-face methods, which are implemented as noop. This is a common trick in generic programming. (ATL programmers will recall the CComFake-CriticalSection used in CCom-Single-Thread-Model and CComMultiThread-Model that does the same trick.)

JNIEnv Smart Pointers

In Listing 6, we solved issues 1 and 2; we no longer have a global variable, and we don't have a timing issue when calling AttachCurrentThread() and Detach-CurrentThread(). However, issue 3 - how to ensure that DetachCurrent-Thread() is called - is still unresolved.

When doing JNI programming, most of the time we're dealing with the JNIEnv* pointer rather than the JavaVM* pointer. The most important characteristic of the JNIEnv pointer is that it has thread affinity (it's only valid in its associated thread). It would be nice if we could get the JNIEnv pointer in an arbitrary context. To achieve this and solve issue 3, JTL provides three smart pointers: simple_env_ ptr, auto_env_ptr, and thread_env_ptr. They're all class templates.

Before talking about these smart pointers, let's review the scenarios in which the JNIEnv pointer is used:

  1. In the launcher thread: In this case the ThreadingModel template parameter should be SingleThread-Model to avoid unnecessary thread synchronization, even though there may be multiple threads.
  2. In multiple threads: AttachCurrent-Thread() and Detach-CurrentThread() will be called only once for each thread.
  3. In multiple threads: AttachCurrent-Thread() and Detach-CurrentThread() can be called multiple times for each thread (see below).
simple_env_ptr, auto_env_ptr and thread_env_ptr are corresponding solutions for scenarios 1, 2, and 3. All these smart pointers are derived from env_ ptr_base, which is a placeholder for JNIEnv* pointer and has other helper functions, such as operator->(), operator!(). By the way, all three JNIEnv smart pointers are not full-blown. For more details about smart pointers, please refer to Alexandrescu's book.

The simple_env_ptr is defined as the following:


template < ... >
class simple_env_ptr : public env_ptr_base {
public:
    //... typedefs for SingletonJVM;
public:
    simple_env_ptr() {
        SingletonJVM::Instance().GetEnv(&env_, jvm_version);
    }
    ~simple_env_ptr() { env_ = 0;}
};

In simple_env_ptr's constructor, the JNIEnv pointer has been initialized by calling GetEnv(). This works because the smart pointer is in the launcher thread. Similarly, auto_env_ptr is defined as the following:


template < ... >
class auto_env_ptr : public env_ptr_base {
public:
    //... typedefs for SingletonJVM;
public:
    auto_env_ptr() {
        SingletonJVM::Instance().AttachCurrentThread(&env_, 0);
    }
    ~auto_env_ptr() {
        if(env_) {
            env_ = 0;
            SingletonJVM::Instance().DetachCurrentThread();
        }
    }
}; // auto_env_ptr

As you have seen, auto_env_ptr's constructor and destructor do the job of attaching and detaching the current thread. Thus we have solved issue 3 and we don't need to worry about the missing detaching call. However, auto_env_ ptr may not work if there is more than one auto_env_ptr instance in the same thread. This is clear in the following snippet:


// auto_env_ptr ptr1, ptr2
ptr1;
// ...other stuff
{
    ptr2;
}
// ... from here all calls on ptr1 may be invalid

In this snippet, when ptr2 is out of scope, its destructor will be called, causing the current thread to detach from the JVM. This may cause problems because ptr1 is still active; however, all its local references may have been freed by the JVM because of the ptr2's detaching call.

To solve this problem, JTL provides thread_env_ptr, which is defined in Listing 8. thread_env_ptr has a counter that's a thread local storage variable. Whenever there is a new thread_env _ptr instance, the counter will be bumped by one. Whenever there is a thread_env_ptr out of scope, the counter will be decreased by one. When the counter is decreased to zero, the current thread will be detached from the JVM.

Putting It All Together

Listing 9 illustrates how to use the JTL JVM invocation. You may provide your own JVMLauncher if the default one does not meet your requirement.

Conclusion

JTL provides complete support for the JVM invocation. It also provides JNIEnv smart pointers to get the JNIEnv* pointer in arbitrary context. Most of the classes in JTL are designed as class templates and they can easily be extended. JTL provides many template parameter implementations as well, which can be used as the default template parameter value under most circumstances.

All comments about JTL and this article are highly appreciated.

More Stories By Stanley Wang

Stanley Wang is a lead software developer at Vcom3D, Inc., and the author of JTL. He received his MS in computer science from the University of Florida. He is interested in system programming, generic programming, and computer graphics.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


@ThingsExpo Stories
Connected devices and the Internet of Things are getting significant momentum in 2014. In his session at Internet of @ThingsExpo, Jim Hunter, Chief Scientist & Technology Evangelist at Greenwave Systems, examined three key elements that together will drive mass adoption of the IoT before the end of 2015. The first element is the recent advent of robust open source protocols (like AllJoyn and WebRTC) that facilitate M2M communication. The second is broad availability of flexible, cost-effective storage designed to handle the massive surge in back-end data in a world where timely analytics is e...
How do APIs and IoT relate? The answer is not as simple as merely adding an API on top of a dumb device, but rather about understanding the architectural patterns for implementing an IoT fabric. There are typically two or three trends: Exposing the device to a management framework Exposing that management framework to a business centric logic Exposing that business layer and data to end users. This last trend is the IoT stack, which involves a new shift in the separation of what stuff happens, where data lives and where the interface lies. For instance, it's a mix of architectural styles ...
The Internet of Things will put IT to its ultimate test by creating infinite new opportunities to digitize products and services, generate and analyze new data to improve customer satisfaction, and discover new ways to gain a competitive advantage across nearly every industry. In order to help corporate business units to capitalize on the rapidly evolving IoT opportunities, IT must stand up to a new set of challenges. In his session at @ThingsExpo, Jeff Kaplan, Managing Director of THINKstrategies, will examine why IT must finally fulfill its role in support of its SBUs or face a new round of...
We are reaching the end of the beginning with WebRTC, and real systems using this technology have begun to appear. One challenge that faces every WebRTC deployment (in some form or another) is identity management. For example, if you have an existing service – possibly built on a variety of different PaaS/SaaS offerings – and you want to add real-time communications you are faced with a challenge relating to user management, authentication, authorization, and validation. Service providers will want to use their existing identities, but these will have credentials already that are (hopefully) i...
Cultural, regulatory, environmental, political and economic (CREPE) conditions over the past decade are creating cross-industry solution spaces that require processes and technologies from both the Internet of Things (IoT), and Data Management and Analytics (DMA). These solution spaces are evolving into Sensor Analytics Ecosystems (SAE) that represent significant new opportunities for organizations of all types. Public Utilities throughout the world, providing electricity, natural gas and water, are pursuing SmartGrid initiatives that represent one of the more mature examples of SAE. We have s...
"Matrix is an ambitious open standard and implementation that's set up to break down the fragmentation problems that exist in IP messaging and VoIP communication," explained John Woolf, Technical Evangelist at Matrix, in this SYS-CON.tv interview at @ThingsExpo, held Nov 4–6, 2014, at the Santa Clara Convention Center in Santa Clara, CA.
The Internet of Things will greatly expand the opportunities for data collection and new business models driven off of that data. In her session at @ThingsExpo, Esmeralda Swartz, CMO of MetraTech, discussed how for this to be effective you not only need to have infrastructure and operational models capable of utilizing this new phenomenon, but increasingly service providers will need to convince a skeptical public to participate. Get ready to show them the money!
One of the biggest challenges when developing connected devices is identifying user value and delivering it through successful user experiences. In his session at Internet of @ThingsExpo, Mike Kuniavsky, Principal Scientist, Innovation Services at PARC, described an IoT-specific approach to user experience design that combines approaches from interaction design, industrial design and service design to create experiences that go beyond simple connected gadgets to create lasting, multi-device experiences grounded in people's real needs and desires.
P2P RTC will impact the landscape of communications, shifting from traditional telephony style communications models to OTT (Over-The-Top) cloud assisted & PaaS (Platform as a Service) communication services. The P2P shift will impact many areas of our lives, from mobile communication, human interactive web services, RTC and telephony infrastructure, user federation, security and privacy implications, business costs, and scalability. In his session at @ThingsExpo, Robin Raymond, Chief Architect at Hookflash, will walk through the shifting landscape of traditional telephone and voice services ...
Scott Jenson leads a project called The Physical Web within the Chrome team at Google. Project members are working to take the scalability and openness of the web and use it to talk to the exponentially exploding range of smart devices. Nearly every company today working on the IoT comes up with the same basic solution: use my server and you'll be fine. But if we really believe there will be trillions of these devices, that just can't scale. We need a system that is open a scalable and by using the URL as a basic building block, we open this up and get the same resilience that the web enjoys.
The Internet of Things is tied together with a thin strand that is known as time. Coincidentally, at the core of nearly all data analytics is a timestamp. When working with time series data there are a few core principles that everyone should consider, especially across datasets where time is the common boundary. In his session at Internet of @ThingsExpo, Jim Scott, Director of Enterprise Strategy & Architecture at MapR Technologies, discussed single-value, geo-spatial, and log time series data. By focusing on enterprise applications and the data center, he will use OpenTSDB as an example t...
The Domain Name Service (DNS) is one of the most important components in networking infrastructure, enabling users and services to access applications by translating URLs (names) into IP addresses (numbers). Because every icon and URL and all embedded content on a website requires a DNS lookup loading complex sites necessitates hundreds of DNS queries. In addition, as more internet-enabled ‘Things' get connected, people will rely on DNS to name and find their fridges, toasters and toilets. According to a recent IDG Research Services Survey this rate of traffic will only grow. What's driving t...
Enthusiasm for the Internet of Things has reached an all-time high. In 2013 alone, venture capitalists spent more than $1 billion dollars investing in the IoT space. With "smart" appliances and devices, IoT covers wearable smart devices, cloud services to hardware companies. Nest, a Google company, detects temperatures inside homes and automatically adjusts it by tracking its user's habit. These technologies are quickly developing and with it come challenges such as bridging infrastructure gaps, abiding by privacy concerns and making the concept a reality. These challenges can't be addressed w...
Explosive growth in connected devices. Enormous amounts of data for collection and analysis. Critical use of data for split-second decision making and actionable information. All three are factors in making the Internet of Things a reality. Yet, any one factor would have an IT organization pondering its infrastructure strategy. How should your organization enhance its IT framework to enable an Internet of Things implementation? In his session at Internet of @ThingsExpo, James Kirkland, Chief Architect for the Internet of Things and Intelligent Systems at Red Hat, described how to revolutioniz...
Bit6 today issued a challenge to the technology community implementing Web Real Time Communication (WebRTC). To leap beyond WebRTC’s significant limitations and fully leverage its underlying value to accelerate innovation, application developers need to consider the entire communications ecosystem.
The definition of IoT is not new, in fact it’s been around for over a decade. What has changed is the public's awareness that the technology we use on a daily basis has caught up on the vision of an always on, always connected world. If you look into the details of what comprises the IoT, you’ll see that it includes everything from cloud computing, Big Data analytics, “Things,” Web communication, applications, network, storage, etc. It is essentially including everything connected online from hardware to software, or as we like to say, it’s an Internet of many different things. The difference ...
Cloud Expo 2014 TV commercials will feature @ThingsExpo, which was launched in June, 2014 at New York City's Javits Center as the largest 'Internet of Things' event in the world.
SYS-CON Events announced today that Windstream, a leading provider of advanced network and cloud communications, has been named “Silver Sponsor” of SYS-CON's 16th International Cloud Expo®, which will take place on June 9–11, 2015, at the Javits Center in New York, NY. Windstream (Nasdaq: WIN), a FORTUNE 500 and S&P 500 company, is a leading provider of advanced network communications, including cloud computing and managed services, to businesses nationwide. The company also offers broadband, phone and digital TV services to consumers primarily in rural areas.
"There is a natural synchronization between the business models, the IoT is there to support ,” explained Brendan O'Brien, Co-founder and Chief Architect of Aria Systems, in this SYS-CON.tv interview at the 15th International Cloud Expo®, held Nov 4–6, 2014, at the Santa Clara Convention Center in Santa Clara, CA.
The major cloud platforms defy a simple, side-by-side analysis. Each of the major IaaS public-cloud platforms offers their own unique strengths and functionality. Options for on-site private cloud are diverse as well, and must be designed and deployed while taking existing legacy architecture and infrastructure into account. Then the reality is that most enterprises are embarking on a hybrid cloud strategy and programs. In this Power Panel at 15th Cloud Expo (http://www.CloudComputingExpo.com), moderated by Ashar Baig, Research Director, Cloud, at Gigaom Research, Nate Gordon, Director of T...