docroot/doc/implementation.html

<?xml version="1.0" encoding="UTF-8" ?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en">
<head>
  <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
  <link rel="stylesheet" href="resources/doc.css" charset="UTF-8" type="text/css" />
  <link rel="stylesheet" href="../coverage/jacoco-resources/prettify.css" charset="UTF-8" type="text/css" />
  <link rel="shortcut icon" href="resources/report.gif" type="image/gif" />
  <script type="text/javascript" src="../coverage/jacoco-resources/prettify.js"></script>
  <title>JaCoCo - Implementation Design</title>
</head>
<body onload="prettyPrint()">

<div class="breadcrumb">
  <a href="../index.html" class="el_report">JaCoCo</a> &gt;
  <a href="index.html" class="el_group">Documentation</a> &gt;
  <span class="el_source">Implementation Design</span>
</div>
<div id="content">

<h1>Implementation Design</h1>

<p>
  This is a unordered list of implementation design decisions. Each topic tries
  to follow this structure:
</p>

<ul>
  <li>Problem statement</li>
  <li>Proposed Solution</li>
  <li>Alternatives and Discussion</li>
</ul>


<h2>Coverage Analysis Mechanism</h2>

<p class="intro">
  Coverage information has to be collected at runtime. For this purpose JaCoCo
  creates instrumented versions of the original class definitions. The
  instrumentation process happens on-the-fly during class loading using so
  called Java agents.
</p>

<p>
  There are several different approaches to collect coverage information. For
  each approach different implementation techniques are known. The following
  diagram gives an overview with the techniques used by JaCoCo highlighted:
</p>

<img src="resources/implementation.png" alt="Coverage Implementation Techniques"/>

<p>
  Byte code instrumentation is very fast, can be implemented in pure Java and
  works with every Java VM. On-the-fly instrumentation with the Java agent
  hook can be added to the JVM without any modification of the target
  application.
</p>

<p>
  The Java agent hook requires at least 1.5 JVMs. Class files compiled with
  debug information (line numbers) allow for source code highlighting. Unluckily
  some Java language constructs get compiled to byte code that produces
  unexpected highlighting results, especially in case of implicitly generated
  code like default constructors or control structures for finally statements.
</p>


<h2>Coverage Agent Isolation</h2>

<p class="intro">
  The Java agent is loaded by the application class loader. Therefore the
  classes of the agent live in the same name space like the application classes
  which can result in clashes especially with the third party library ASM. The
  JoCoCo build therefore moves all agent classes into a unique package.
</p>

<p>
  The JaCoCo build renames all classes contained in the
  <code>jacocoagent.jar</code> into classes with a
  <code>org.jacoco.agent.rt_&lt;randomid&gt;</code> prefix, including the
  required ASM library classes. The identifier is created from a random number.
  As the agent does not provide any API, no one should be affected by this
  renaming. This trick also allows that JaCoCo tests can be verified with
  JaCoCo.
</p>


<h2>Minimal Java Version</h2>

<p class="intro">
  JaCoCo requires Java 1.5.
</p>

<p>
  The Java agent mechanism used for on-the-fly instrumentation became available
  with Java 1.5 VMs. Coding and testing with Java 1.5 language level is more
  efficient, less error-prone &ndash; and more fun than with older versions.
  JaCoCo will still allow to run against Java code compiled for these.
</p>


<h2>Byte Code Manipulation</h2>

<p class="intro">
  Instrumentation requires mechanisms to modify and generate Java byte code.
  JaCoCo uses the ASM library for this purpose internally.
</p>

<p>
  Implementing the Java byte code specification would be an extensive and
  error-prone task. Therefore an existing library should be used. The
  <a href="http://asm.objectweb.org/">ASM</a> library is lightweight, easy to
  use and very efficient in terms of memory and CPU usage. It is actively
  maintained and includes as huge regression test suite. Its simplified BSD
  license is approved by the Eclipse Foundation for usage with EPL products.
</p>

<h2>Java Class Identity</h2>

<p class="intro">
  Each class loaded at runtime needs a unique identity to associate coverage data with.
  JaCoCo creates such identities by a CRC64 hash code of the raw class definition.
</p>

<p>
  In multi-classloader environments the plain name of a class does not
  unambiguously identify a class. For example OSGi allows to use different
  versions of the same class to be loaded within the same VM. In complex
  deployment scenarios the actual version of the test target might be different
  from current development version. A code coverage report should guarantee that
  the presented figures are extracted from a valid test target. A hash code of
  the class definitions allows to differentiate between classes and versions of
  classes. The CRC64 hash computation is simple and fast resulting in a small 64
  bit identifier.
</p>

<p>
  The same class definition might be loaded by class loaders which will result
  in different classes for the Java runtime system. For coverage analysis this
  distinction should be irrelevant. Class definitions might be altered by other
  instrumentation based technologies (e.g. AspectJ). In this case the hash code
  will change and identity gets lost. On the other hand code coverage analysis
  based on classes that have been somehow altered will produce unexpected
  results. The CRC64 code might produce so called <i>collisions</i>, i.e.
  creating the same hash code for two different classes. Although CRC64 is not
  cryptographically strong and collision examples can be easily computed, for
  regular class files the collision probability is very low.
</p>

<h2>Coverage Runtime Dependency</h2>

<p class="intro">
  Instrumented code typically gets a dependency to a coverage runtime which is
  responsible for collecting and storing execution data. JaCoCo uses JRE types
  only in generated instrumentation code.
</p>

<p>
  Making a runtime library available to all instrumented classes can be a
  painful or impossible task in frameworks that use their own class loading
  mechanisms. Since Java 1.6 <code>java.lang.instrument.Instrumentation</code>
  has an API to extends the bootsstrap loader. As our minimum target is Java 1.5
  JaCoCo decouples the instrumented classes and the coverage runtime through
  official JRE API types only. The instrumented classes communicate through the
  <code>Object.equals(Object)</code> method with the runtime. A instrumented
  class can retrieve its probe array instance with the following code. Note
  that only JRE APIs are used:
</p>


<pre class="source lang-java linenums">
Object access = ...                          // Retrieve instance

Object[] args = new Object[3];
args[0] = Long.valueOf(8060044182221863588); // class id
args[1] = "com/example/MyClass";             // class name
args[2] = Integer.valueOf(24);               // probe count

access.equals(args);

boolean[] probes = (boolean[]) args[0];
</pre>

<p>
  The most tricky part takes place in line 1 and is not shown in the snippet
  above. The object instance providing access to the coverage runtime through
  its <code>equals()</code> method has to be obtained. Different approaches have
  been implemented and tested so far:
</p>

<ul>
  <li><b><code>SystemPropertiesRuntime</code></b>: This approach stores the
    object instance under a system property. This solution breaks the contract
    that system properties must only contain <code>java.lang.String</code>
    values and therefore causes trouble in applications that rely on this
    definition (e.g. Ant).</li>
  <li><b><code>LoggerRuntime</code></b>: Here we use a shared
    <code>java.util.logging.Logger</code> and communicate through the logging
    parameter array instead of a <code>equals()</code> method. The coverage
    runtime registers a custom <code>Handler</code> to receive the parameter
    array. This approach might break environments that install their own log
    managers (e.g. Glassfish).</li>
  <li><b><code>URLStreamHandlerRuntime</code></b>: This runtime registers a
    <code>URLStreamHandler</code> for a "jacoco-xxxxx" protocol. Instrumented
    classes open a connection on this protocol. The returned connection object
    is the one that provides access to the coverage runtime through its
    <code>equals()</code> method. However to register the protocol the runtime
    needs to access internal members of the <code>java.net.URL</code> class.</li>
  <li><b><code>ModifiedSystemClassRuntime</code></b>: This approach adds a
    public static field to an existing JRE class through instrumentation. Unlike
    the other methods above this is only possible for environments where a Java
    agent is active.</li>
  <li><b><code>InjectedClassRuntime</code></b>: This approach defines a new class
    using <code>java.lang.invoke.MethodHandles.Lookup.defineClass</code>
    introduced in Java 9.</li>
</ul>

<p>
  Starting from version 0.8.3 JaCoCo Java agent implementation uses the
  <code>InjectedClassRuntime</code> to define new class in bootstrap class
  loader when running on JRE 9 and higher, otherwise uses
  <code>ModifiedSystemClassRuntime</code> to add field to an existing JRE class.
  Starting from version 0.8.0 field is added to the class
  <code>java.lang.UnknownError</code>, versions 0.5.0 - 0.7.9 were adding field
  to the class <code>java.util.UUID</code>, having bigger chance of conflict
  with other agents.
</p>


<h2>Memory Usage</h2>

<p class="intro">
  Coverage analysis for huge projects with several thousand classes or hundred
  thousand lines of code should be possible. To allow this with reasonable
  memory usage the coverage analysis is based on streaming patterns and
  "depth first" traversals.
</p>

<p>
  The complete data tree of a huge coverage report is too big to fit into a
  reasonable heap memory configuration. Therefore the coverage analysis and
  report generation is implemented as "depth first" traversals. Which means that
  at any point in time only the following data has to be held in working memory:
</p>

<ul>
  <li>A single class which is currently processed.</li>
  <li>The summary information of all parents of this class (package, groups).</li>
</ul>

<h2>Java Element Identifiers</h2>

<p class="intro">
  The Java language and the Java VM use different String representation formats
  for Java elements. For example while a type reference in Java reads like
  <code>java.lang.Object</code>, the VM references the same type as
  <code>Ljava/lang/Object;</code>. The JaCoCo API is based on VM identifiers only.
</p>

<p>
  Using VM identifiers directly does not cause any transformation overhead at
  runtime. There are several programming languages based on the Java VM that
  might use different notations. Specific transformations should therefore only
  happen at the user interface level, for example during report generation.
</p>

<h2>Modularization of the JaCoCo implementation</h2>

<p class="intro">
  JaCoCo is implemented in several modules providing different functionality.
  These modules are provided as OSGi bundles with proper manifest files. But
  there are no dependencies on OSGi itself.
</p>

<p>
  Using OSGi bundles allows well defined dependencies at development time and
  at runtime in OSGi containers. As there are no dependencies on OSGi, the
  bundles can also be used like regular JAR files.
</p>

</div>
<div class="footer">
  <span class="right"><a href="@jacoco.home.url@">JaCoCo</a> @qualified.bundle.version@</span>
  <a href="license.html">Copyright</a> &copy; @copyright.years@ Mountainminds GmbH &amp; Co. KG and Contributors
</div>

</body>
</html>