Getting started - Beginning

How to use the Java Scilab binding?

Description

The goal is this module is to allow the load and interaction on Java objects and datatypes.

Basic

Before starting, it is convenient to know the most important functions and what they do. These frequently used functions are the following Scilab functions:

jimport: Import a Java class
jinvoke: Invoke a method of a Java object

jimport is the function that mirrors the functionality of the java statement 'import', and loads the specified class definition/template of a given class in memory. When loaded, this definition is used to accesses static methods/members, and create new objects.

jinvoke is a function that calls (invokes) a specified method on a java class or object. This invoke has an optional set of parameters that must coincide with the actual member signature. This means you must pass the same number of arguments, and these arguments must have the correct type.

Example 1: Creating a basic class and calling a simple method

In this first example, the three basic pillars of working with Java are treated. The first is to load a class, then an instance is constructed and finally the invocation or calling of one of this methods or members.

Consider a basic class as presented in the example HelloWorld. It has a default constructor generating a message upon construction and one public method which also shows a message when it has been called. This class should now be compiled into java byte-code. When developing your own code, then this section is usually handled by your IDE (integrated development environment). If you plan to use external libraries, these will available in precompiled form (packed in a JAR).

// Save under the name HelloWorld.java
package com.foo;
public class HelloWorld {
   public HelloWorld() {
      System.err.println("HelloWorld constructed!");
  }

  public void message() {
      System.err.println("Hello world!");
  }
}

// How to compile the Java code from Scilab
javacode=mgetl(fullfile(TMPDIR, 'HelloWorld.java'));
jcompile("com.foo.HelloWorld",javacode);

Once the compiled version of this Java class exists, we can start Scilab and try to get Scilab to show us the various messages. Now the HelloWorld class can be imported into our workspace. This is done using the already mentioned jimport:

                -->jimport com.foo.HelloWorld

                -->HelloWorld
                HelloWorld  =

                class com.foo.HelloWorld

                -->whos -name HelloWorld
                Name                     Type           Size           Bytes

                HelloWorld               _EClass        ?              168

Upon competition, an additional variable named HelloWorld has been created. This is equivalent to a Class object in java. From this class object, new objects of the HelloWorld type can be created.

Creating such an object instance can be done by invoking new on the class definition. The arguments to this function are the parameters that are delegated to the Java constructor. The result of this operation is a new Java object reference which can be stored in a variable for later use.

                -->object = HelloWorld.new();
                HelloWorld constructed!

                -->object
                object  =

                com.foo.HelloWorld@49aacd5f

                -->whos -name object
                Name                     Type           Size           Bytes

                object                   _EObj          ?              160

What one sees when the new operator is called on the JClass, it transparently invokes the Java constructor, and our "HelloWorld constructed!" message appears. The resulting HelloWorld object is stored in the "object" variable. When the variable name is reentered in the command line, the details of its reference are shown. This message can be customized by overriding the toString method in the HelloWorld class.

Now that a specific HelloWorld object has been created, one can try to call the public method that has been declared; HelloWorld\#message(). The same natural technique as with new can then be applied to invoke the method:

                -->object.message();
                Hello world!

The dot operator (dot between object and message) is actually a handy shortcut and expands the following snippet of Scilab code. The use of this shortcut makes it simpler and cleaner to invoke methods/get member variables.

                -->jinvoke(object, 'message');
                Hello world!

Example 2: Exchanging Scilab and Java primitives

This example treats the way you can exchange primitive data types and strings between Scilab and Java. We will be passing various types of objects between these two languages.

For this an example class (see Class Inspector) has been defined that takes and returns objects. There are two methods defined. The first takes a double does some arithmetic and returns a result: Inspector#eval(double). The other methods takes any object, shows some basic information and returns it: Inspector#inspect(Object).

// Save under the name Inspector.java
package com.foo;
public class Inspector {
	public double eval(double x) {
		return x / 2.;
	}

	public Object inspect(Object prototype) {
		System.err.println("Inspecting: '" + prototype + "'");
		System.err.println("Class: " + prototype.getClass().getSimpleName());
		return prototype;
	}
}

As in the previous example, this code must be compiled to Java byte-code before it can be used directly.

// How to compile the Java code from Scilab
javacode= mgetl(fullfile(TMPDIR, 'Inspector.java'));
jcompile("com.foo.Inspector",javacode);

To start, import the Inspector class and create an Inspector object:

            -->jimport('com.foo.Inspector');

            -->myInspector = Inspector.new()
            myInspector  =

            com.foo.Inspector@2a788315

Now information between the two systems can be passing along. When passing along any of the Scilab data types into Java, it is automatically wrapped (see jwrap) to its Java equivalent. First, an example using the most used data type in Scilab; reals (constant) is given. When passing along a real, this type gets automatically mapped to the Scilab type double. Let's try;

            -->result = myInspector.eval(12.5)
            result  =

            6.25

            -->result * 2
            ans  =

            12.5

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   constant       1 by 1         24

The automatic convert is controlled by the jautoUnwrap function. Without this function, all the conversions have to be done explicitly.

            -->result = myInspector.eval(12.5)
            result  =

            6.25

            -->result * 2
            ans  =

            null

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   _EObj          ?              160

The result that is returned seems to be correct at first glance ($12.5/2=6.25$). However upon closer inspection, notice that what is returned from our function call is not a number. What we received is another Java object (Double in this case). To be able to use the given data in again in Scilab, the already mentioned junwrap functionality can be used if the jautoUnwrap is not set to true. This transforms Java types back to there equivalent Scilab form. From then onward we have a normal number again:

            -->result = junwrap(result)
            result  =

            6.25

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   constant       1 by 1         24

            -->result * 2
            ans  =

            12.5

From this example is clear how doubles get automatically transformed into a Double, which is used by the Java VM and returned back. When calling junwrap on the returned variable, it is transformed back into a native Scilab type. But how do other types work? Let's inspect several of the other primitive types;

            -->jautoUnwrap(%f) // Make sure we disable the auto Unwrap

            -->result = myInspector.inspect("Hello world!");
            Inspecting: 'Hello world!'
            Class: String

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   _EObj          ?              160

            -->result = junwrap(result)
            result  =

            Hello world!

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   string         1 by 1         72

            // An Integer
            -->result = myInspector.inspect(int32(150));
            Inspecting: '150'
            Class: Integer

            -->result = junwrap(result)
            result  =

            150

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   int32          1 by 1         40

            // A boolean
            -->result = myInspector.inspect(%t);
            Inspecting: 'true'
            Class: Boolean

            -->result = junwrap(result)
            result  =

            T

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   boolean        1 by 1         16

As can be seen, all relevant data types are can be transformed transparently between Scilab and Java type. However this also extends without any additional effort to matrices;

            --> jautoUnwrap(%t) // Make sure we come back in the default mode where Scilab auto unwrap all calls

            -->result = myInspector.inspect(1:5)
            Inspecting: '[D@b05236'
            Class: double[]
            result  =

            1.    2.    3.    4.    5.

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   constant       1 by 5         56

            -->result = myInspector.inspect(testmatrix('magi',3))
            Inspecting: '[[D@11d13272'
            Class: double[][]
            result  =

            8.    1.    6.
            3.    5.    7.
            4.    9.    2.

            -->whos -name result
            Name                     Type           Size           Bytes

            result                   constant       3 by 3         88

When looking at the class of these wrapped matrices, it is clear that Java stores them as arrays of the appropriate size. When working with 2D matrices, the data in these equivalent Java arrays can be stored in are column-major (default) or row-major mode. In column-major mode, the first array contains a pointer to each of the columns. Whereas in row-major mode, the first array contains the pointers to each row of data. For more information see jautoTranspose.

History

Version	Description
5.5.0	Function introduced. Based on the 'JIMS' module. The main difference in the behavior compared to the JIMS module is that jautoUnwrap is enabled by default.

Report an issue
<< Java from Scilab	Java from Scilab	Getting started - Second step >>

Copyright (c) 2022-2024 (Dassault Systèmes)
Copyright (c) 2017-2022 (ESI Group)
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors

Last updated:
Mon Feb 12 19:58:46 CET 2018