Java

Runtime library

Following most other runtimes example, everything revolves around two basic classes:

• KaitaiStruct — a common superclass for all classes that represent user types in KS. Java implementation is very limited and basically only keeps _io member of type KaitaiStream and provides a getter for it.
• KaitaiStream — a useful abstraction of seekable input stream that can be read with Kaitai Struct stream API (i.e. methods like readU4le(). Internally, it uses either a ByteArrayInputStream or RandomAccessFile to read either a in-memory byte buffer or a local file.

Types

KS tries to follow mandatory and recommended Java practices as close as possible.

All user types are mapped 1-to-1 to relevant Java classes. Class names would be represented in upper camel case (i.e. an_example_class => AnExampleClass). Nested types are mapped to nested classes, i.e. for nested types like this:

meta:
  id: parent
# ...
types:
  child:
    # ...
    types:
      grandchild:
        # ...

one can expect to get the following class structure:

public class Parent extends KaitaiStruct {
    public static class Child extends KaitaiStruct {
        public static class GrandChild extends KaitaiStruct {
        }
    }
}    

Every generated class will have 3 constructors and a static factory method:

public AnExampleClass(KaitaiStream _io)
public AnExampleClass(KaitaiStream _io, KaitaiStruct _parent)
public AnExampleClass(KaitaiStream _io, KaitaiStruct _parent, AnExampleClass _root)
public static AnExampleClass fromFile(String fileName)

Most straightforward end-users

Attributes and instances

All attributes and instance names use lower camel case (i.e. an_example_attribute => anExampleAttribute). For all attributes, a relevant getter method will be generated, so an attribute can be accessed outside of class like classInstance.anExampleAttribute().

Instances

TODO

Primitive type mapping

There are several things of note that influence mapping KS types to Java types:

• There are no support for unsigned integer types in Java. In some cases it's no big deal, but some use cases (for example, comparison or bit shifts) may be severely hindered by that issue. KS tries to make up for that fact by using larger signed types where that's possible and reasonable to do. Where it's not possible (i.e. 64-bit unsigned integers — u8), KS would use signed long type.
• Java has 2 types for every numeric type: "primitive" type (i.e. int) and "reference" type (i.e. Integer) — the latter being a full-featured object that can have null assigned to it and stored in collections. It's not practical to use reference types everywhere, so KS makes use of them only in the following situations:
  • when data type is used as part of a collection
  • when it's possible that a particular attribute / instance will be unassigned (i.e. because of if expression) — null will be returned in this case

The overall primitive type mapping goes as follows:

type	Java primitive type	Java reference type
no type	byte[]	byte[]
u1	int	Integer
u2	int	Integer
u4	long	Long
u8	long	Long
s1	byte	Byte
s2	short	Short
s4	int	Integer
s8	long	Long
str, strz	String	String

Array types

All repetitions in Java are translated to ArrayList<~>

String encoding

Encoding a stream of bytes into a String is done with the standard Java API: String method constructor

Enums

TODO