Skip to content

This Maven plugin converts Java Bytecode binary files to EOLANG programs that use the "opcode" atom

License

Notifications You must be signed in to change notification settings

objectionary/jeo-maven-plugin

Repository files navigation

logo

Maven Central Javadoc License Hits-of-Code Lines of code codecov

jeo stands for "Java EOlang Optimizations". jeo-maven-plugin is a Maven plugin dedicated to optimizing Java bytecode. The process involves translating the Java bytecode into the EOlang programming language. Utilizing the optimization steps provided by EOlang, the original code undergoes an enhancement process. Upon completion, the optimized EOlang program is translated back to Java bytecode, achieving efficient and optimized performance.

How to use

The plugin can be run using several approaches but for all of them you need at least Maven 3.1.+ and Java 8+. The plugin can convert compiled classes into EOlang by using the disassemble goal. The assemble goal can convert EOlang back into bytecode. The default phase for the plugin is process-classes. If you are a developer of optimizations in EOlang you probably need to use the both goals in the following order:

  • disassemble create EOlang files in the target/generated-sources directory.
  • Provide your optimizations are applied to the EOlang files in the target/generated-sources directory.
  • assemble scans the target/generated-sources directory for EOlang files and converts them back to Java bytecode.

More details about plugin usage you can find in our Maven site.

Invoke the plugin from the command line

You can run the plugin directly from the command line using the following commands:

mvn jeo:disassemble

or

mvn jeo:assemble

Invoke the plugin from the Maven lifecycle

You can run the plugin from the Maven lifecycle by adding the following configuration to your pom.xml file:

<build>
  <plugins>
    <plugin>
      <groupId>org.eolang</groupId>
      <artifactId>jeo-maven-plugin</artifactId>
      <version>0.6.26</version>
      <executions>
        <execution>
          <id>bytecode-to-eo</id>
          <phase>process-classes</phase>
          <goals>
            <goal>disassemble</goal>
          </goals>
        </execution>
        <execution>
          <id>eo-to-bytecode</id>
          <phase>process-classes</phase>
          <goals>
            <goal>assemble</goal>
          </goals>
        </execution>
      </executions>
    </plugin>
  </plugins>
</build>

Include debug information

In order to include debug information in the generated EO files, you can set debug option.

<configuration>
  <mode>debug</mode>
</configuration>

This option will add line numbers and local variable names to the EO files together with their corresponding labels.

Disable bytecode verification

Each time the plugin converts EO back to bytecode, it verifies it. If the verification fails, the build also fails. You can disable this verification by setting the skipVerification parameter to true:

<configuration>
  <skipVerification>true</skipVerification>
</configuration>

At times, it might be beneficial to generate intentionally flawed bytecode.

Transformation method

The plugin can transform Java bytecode into EO and back. Usually, the plugin transforms each bytecode class file into a separate EO file, maintaining a one-to-one relationship. If the Java class has name Foo.class, the EO file will have Foo.eo (and Foo.xmir for the XMIR representation of the EO file).

Classes

The first high-level transformation is the conversion of the bytecode class into <program> and <objects><o name='Classname'/></objects> XMIR elements. For example, consider the following Java class:

public class Foo {
}

It will be transformed into the following EO:

[] > j$Foo
  33 > access
  "java/lang/Object" > supername
  * > interfaces

access(class access modifiers like public, static, final and others), supername (parent class), and interfaces (tuple of implemented interfaces) are attributes of the class element that retain the information necessary for the reverse transformation.

The j$* prefix is employed to prevent name conflicts with EO keywords. This same prefix is utilized for all EO elements generated from Java bytecode.

By the way, the XMIR representation of that EO file will be:

<program>
  <objects>
    <o name="j$Foo">
      <attribute name="access" value="33"/>
      <attribute name="supername" value="java/lang/Object"/>
      <attribute name="interfaces" value="*"/>
    </o>
  </objects>
</program>

Methods

The second high-level transformation involves converting the bytecode method into EO. For example, consider the following Java method:

public class Bar {
    public void foo() {
        return;
    }
}

It will be transformed into the following EO:

[] > j$Bar
  33 > access
  "java/lang/Object" > supername
  * > interfaces
  [] > j$foo
    1 > access
    "()V" > descriptor
    * > exceptions
    seq > @
      tuple
        opcode > RETURN-1
          177

Each method is a child of the class element and contains bytecode attributes such as access (access modifiers), descriptor (method descriptor), and exceptions (a tuple of declared exceptions). Additionally, it includes the seq element containing the sequence of bytecode instructions.

It's worth mentioning that Java constructors are also treated as methods with the name new. For instance, consider the following Java constructor:

public class Bar {
    public Bar() {
    }
}

It will be transformed into the following EO:

[] > j$Bar
  33 > access
  "java/lang/Object" > supername
  * > interfaces
  [] > new
    1 > access
    "()V" > descriptor
    * > exceptions
    seq > @
      tuple
        // list of instructions

Instructions

Each method and constructor contains a sequence of instructions, with each instruction represented by either a opcode or a label. For example, consider the following Java method:

public class Bar {
    public int foo(int x) {
        if (x < 0) {
            return 1;
        }
        return 2;
    }
}

It will have the following set of instructions after compilation (as shown by javap -v Bar output):

0: iload_1
1: ifle          6
4: iconst_1
5: ireturn
6: iconst_2
7: ireturn

After the transformation provided by jeo, the content of the foo method in EO will look like:

seq > @
  tuple
    label
      "67b715c8-7d74-413a-9bba-f6920c8ba68b"
    opcode > ILOAD-E
       21
       1
    opcode > IFLE-F
      158
      label
        "c361c429-6c81-4b11-9b97-0cbb6e96a2f9"
    opcode > ICONST_1-10
      4
    opcode > IRETURN-11
      172
    label
      "c361c429-6c81-4b11-9b97-0cbb6e96a2f9"
    opcode > ICONST_2-12
      5
    opcode > IRETURN-13
      172
    label
      "8f341f7f-e357-4a78-b604-bcaae28e3c1f"

Opcode

From the example above (refer to the Methods section), you can observe that each opcode is represented by the opcode object. Each opcode object includes a name, a numerical argument, and optional operand arguments. For instance, the iload opcode has the following EO representation

opcode > ILOAD-E
  21
  1

Here, ILOAD-E is the opcode's name, 21 is its number according to the Java specification, and 1 represents the opcode argument, indicating the 'local variable with index 1' in this context.

ILOAD-E is simply a name assigned to an opcode object. Since it serves a descriptive purpose and isn't relied upon during transformations, you have flexibility to modify these names as needed when making changes to the original jeo output (as jeo doesn't utilize them during parsing.)

Also, it's worth mentioning that an opcode might not have operand arguments, as is the case with the IRETURN opcode:

opcode > IRETURN-11
  172

Alternatively, the opcode argument might be a label object, as seen in the IFLE instruction:

opcode > IFLE-F
  158
  label
    "c361c429-6c81-4b11-9b97-0cbb6e96a2f9"

In this instance, the IFLE opcode has precisely one operand, which is a label.

Labels

Labels serve as markers or references indicating specific points in the code:

  1. They might mark the entry- and exit-points of a method for debugging purposes.
  2. They provide jump points in the code, such as for if and for statements. The example of using labels is in conjunction with the goto instruction:
opcode > GOTO-1
  167
  label "dbe5a680-4814-4b19-a8e6-15c3c2db3a83"
opcode > ALOAD-2 // skiped by goto
  25             // skiped by goto
  1              // skiped by goto
label "dbe5a680-4814-4b19-a8e6-15c3c2db3a83"
opcode > RETURN-3
  177
  1. Labels can also be used for exception handling.

Of course, this isn't an exhaustive list of label usages.

What is more important, many labels are as crucial as opcodes themselves, and if subsequent transformations lose these labels, the logic of the program might be compromised. Therefore, it is extremely important to preserve most of the labels. However, it's worth noting that you can omit certain labels used solely for debugging purposes when generating your own classes. For instance, you can omit labels at the start and end of a method.

How to Build the Plugin

To build the plugin from the source code, you need to clone the repository and run the following command:

$ mvn clean install -Pqulice,long

Pay attention to the qulice profile, which activates the static analysis tools. The long profile is optional and runs the full test suite, including long-running integration tests.

How to Contribute

Fork repository, make changes, then send us a pull request. We will review your changes and apply them to the master branch shortly, provided they don't violate our quality standards. To avoid frustration, before sending us your pull request please run full Maven build:

$ mvn clean install -Pqulice

You will need Maven 3.3+ and Java 8+ installed.