IDL to Java 5 Mapping

1
IDL to Java 5 Mapping
2
Perspective

• The views expressed in this presentation are mine
  (but not necessarily mine alone) ?
• The IDL to Java Mappingsneed updating
• Why?...The current mappings are
• Clunky
• Inefficient
• Non-intuitive

3
Specifics

• Which version?
• IDL to Java Language Mapping Specification,
  August 2002, Version 1.2, The Object Management
  Group
• http//www.omg.org/cgi-bin/doc?formal/02-08-05.pdf
  

4
Where are the problems?
Stuff That's Missing...
5
enum

• This one seems straight forward but
• Currently
• enum maps to a public class with static values
  for the enumerates and their integer (serialized)
  representation
• Proposed
• enum maps to a Java enum
• Helper method would provide a conversion from the
  integer representation for deserialization and
  the standard ordinal() method would provide the
  integer for serialization

6
const

• Currently Given the following IDL
• const double PI 3.141592
• const double e 2.718282
• Yields the following Java
• public interface PI double value 3.141592
• public interface e double value 2.718282
• Proposed Given the following IDL
• const double PI 3.141592
• const double e 2.718282
• Would yield the following Java
• public final class module_nameConstants
• public final static double PI 3.141592
• public final static double e 2.718282
• This would allow us to use Java 5s static import
  to gain access to all of the constants for a
  given module with one fell swoop

7
sequence (1 of 3)

• Currently Given the following IDL
• sequence previousOrderIds
• Yields the following Java
• public int previousOrderIds
• The are a couple of problems with this mapping
• Native Java arrays are not resizeable
• Semantics can be improved by using Java 5s
  generics
• No possible way to control memory management (see
  next slide)

8
sequence (2 of 3)

• So what if native Java arrays are not
  resizeable?!
• This means that the clients of the sequence
  mapping will have to write the code below to
  increase the size of a sequence
• int oldSize previousOrderIds.length
• // allocate storage for the new size (larger in
  this case)
• int previousOrderIds new intoldSize1
• // copy over the original data
• System.arraycopy(foo.previousOrderIds, 0,
• previousOrderIds, 0, oldSize)
• // reset the reference to the new int array
• foo.previousOrderIds previousOrderIds
• // add the order id
• foo.previousOrderIdsoldSize someOrderId

9
sequence (3 of 3)

• That seems like a lot of code with plenty of
  opportunity to introduce a bug
• What if we changed the mapping from a native
  array to a generic such as
• public List previousOrderIds
• The code would now look like this
• // add the new order id
• foo.previousOrderIds.add(someOrderId)
• Now thats better! ?
• Additionally, real-time implementations could
  choose to pre-allocate a maximally-sized List
  object!

10
string and wstring (1 of 2)

• Currently IDL string and wstring map to a Java
  String object.
• Most of the time, this is perfect!
• But, if you want to use IDL string or wstring
  attributes in a NoHeapRealTime Thread (RTSJ) then
  there may be some issues
• The problem is that memory must be allocated when
  a String is modified since all Strings are
  immutable
• That doesnt necessarily work well with NHRT

11
string and wstring (2 of 2)

• Alternate Mapping Possibilities
• java.lang.StringBuffer
• May be too heavyweight since it has all
  synchronized methods
• Inconvenient when doing character i/o since it
  requires calling toString() which, in turn,
  allocates memory
• java.lang.StringBuilder
• Ligherweight (no synchronization)
• Same Character i/o issue as StringBuffer
• java.lang.CharSequence
• Provides a read-only interface to the characters
  but could be useful
• java.nio.CharBuffer
• This class is intended more for heavy-duty
  character i/o but should be considered

12
Stuff Thats Missing

• Generics
• Bounded values
• Wouldnt it be nice to be able to declare
• double latitude
• Valued enumerates
• enum StatusKind NONE 0, ALL 0xFFFFFFFF,
  INCOMPATIBLE_QOS 1, etc.
• Or add an enumset
• Others???