Java integration planning

This information is for Ioke E+

Java classes can be accessed by their full name, separated with colon instead of dots:

h = java:util:HashMap new

You can invoke methods as expected, although the rules for this are a bit intricate. Specifically, the rules are that when calling methods that take any kind of primitive arguments (most of the types in java.lang), these values will be unwrapped. That means for a method taking an int, you can send in a wrapped Java Integer, but you can also send in an Ioke Number Rational, and it will be unwrapped correctly. You can send Text and Symbol to methods taking Strings.

If you send in data to a method taking Object, you will not get the expected result in all cases. Specifically, in these cases Ioke will unwrap wrapped Java objects, but will not coerce the primitive types. So doing h put("foo", "bar") will not coerce Text into java.util.String. However, if you have a java.lang.String that has been modified from the Ioke side - adding cells for example, then that String will be unwrapped before sent to the method call.

The return value from a Java invocation will never be modified, except that null will always be changed into nil. This means that if you wrap a Java object, do some modifications to it, and then let it go through a Java call, what you will get back is probably not the wrapped version of that object. Of course, later on if you try to call a method on the object, you will still use the wrapped data.

The general rule for overloading is that overloaded methods will be sorted from the most specific to the most general.

Casts

Ioke Java integration supporting casts in the same way as Java does. The main difference is that the cast will be evaluated at runtime, not compile time. The name to cast to will be unevaluated. Valid names are either Object, String, any Java class name where dots are separated with colons, and a few special ones to handle primitive values. These names are int, integer, long, short, char, character, float, double and boolean. Casting looks like this:

x = Foo new
x bar((short) 123, (org:something:Interface)anObj)

Class methods

Methods that belong to the Java class java.lang.Class are treated a bit differently from the way other Java methods work. The reason for this is a bit intricate to explain, but basically comes down to the dichotomy between the way mimicking works in Ioke, and the way inheritance works in Java.

The easiest way to explain this might be to take a look at java.util.ArrayList. In Java the hierarchy looks like this (somewhat simplified)

instance >> ArrayList -> AbstractList -> AbstractCollection -> Object
                      -> Serializable
                      -> Cloneable
                      -> Iterable
                      -> Collection
                      -> List
                      -> RandomAccess

Note here that all capitalized words are names of classes. The Class itself is not part of the inheritance structure, because Java has two namespaces. In Ioke, the above hierarchy basically ends up looking like this:

instance -> ArrayList -> Class
                      -> AbstractList -> Class
                                      -> AbstractCollection -> Class
                                                            -> Object -> Class
                      -> Serializable -> Class
                      -> Cloneable -> Class
                      -> Iterable -> Class
                      -> Collection -> Class
                      -> List -> Class
                      -> RandomAccess -> Class

This is slightly simplified too, but the basic idea is that Class has to be part of the same inheritance chain, because there only exists one namespace in Ioke.

And that is why there is a need to have these methods named differently. So, for example, say that you want to invoke the toString method on java:util:HashMap. To do that, you do it like this: java:util:HashMap class:toString. Note that you can actually use the class: methods on instances of a class too. The result will be the same as if the receiver had been that class instead of the instance.

Fields

Accessing Java fields can be done by prepending the field name with field:. The exact mechanics of this is that a Java field will result in one or two Ioke methods. That means that the fields are not represented as cells directly on the object. If the field is not final, a setter will be also be generated for it. The setter follows the same rules as invocations of regular methods, with regards to casting and choice of unpacking of arguments.

There is a slight gotcha with this scheme. If you try to set a field that is final, you will end up overwriting the accessor for that field - so it's important to be really careful to not set final fields. In the long run, it might be an idea to implement a setter that signals a condition when this happens, but doing that gives the impression that you can set the value - since the setter is there.

Ioke doesn't care if a field is private or protected. All fields are accessible.

A small example - say that you have a java object in foo. This Java object has two fields called oneThing and anotherThing. Then you can work with it like this:

foo field:oneThing println
foo field:anotherThing = "Bar"
System out println(foo field:anotherThing)

Note that static fields work the same way - so you can access the Java streams directly like this, for example:

java:lang:System field:out println("some text")

Importing

Ioke can import classes just as Java can. The way to do it looks a bit different, though. Also, in Ioke an import only means that a local name will be assigned to a class. It will not necessarily be globally visible, unless you import in a global scope. It's also important to keep in mind that imports will not happen if they might shadow or overwrite a name. If you want that, you will have to do a manual assignment instead.

There are three ways of using the method import. The first one is simply to import one class with the same name as it already has:

;; these are all equivalent
HashMap = java:util:HashMap
import(java:util:HashMap)
import java:util:HashMap

;; you can also import directly into a scope
foo = Origin mimic
foo import(java:util:LinkedHashMap)
foo LinkedHashMap new

Note that you can only give one class to import when using the above way.

The second way to import is when you want to rename the imports. This allow you to import several classes from different packages and also rename them at the same time. This is done using keyword arguments to import:

;; these are all equivalent
Foo = java:util:HashMap
Bar = java:lang:String
import(Foo: java:util:HashMap, Bar: java:lang:String)

;; you can do the same in a new scope:
foo = Origin mimic
foo import(Foo: java:util:HashMap, Bar: java:lang:String)
foo Foo new

The third way allow you to import several classes from the same package. Note that to use this way, you need to provide at least one class name, in addition to the package name:

;; these are the same as the following import
HashMap = java:util:HashMap
LinkedList = java:util:LinkedList
ArrayList = java:util:ArrayList
import(:java:util, :HashMap, :LinkedList, :ArrayList)

;; and just as above, you can do it in a scope too
foo = Origin mimic
foo import(:java:util, :HashMap, :LinkedList, :ArrayList)
foo LinkedList new