POF question

Question

Hi,
Does Portable Object Format (POF) enable me to write objects in Java and have a c++ client that can read them and vice versa?
Is there an example of this in the documentation? Thanks!
Update:
Okay, I think I have answered my own question (but I have more questions, see below). Yes, data can be written from a Java client to Coherence, and read from a C++ client (or a .NET client).
Question 1: In Java, the ConfigurablePofContext must be used to map the numeric identifier to a Java class name. Next, the cluster must be configured to use the ConfigurablePofContext by adding a <serializer> element to the configuration file. Is there a similar config file for C++ or is this all accomplished using the registration methods?:
COH_REGISTER_MANAGED_CLASS(ID, TYPE) - for use with Managed<T>
COH_REGISTER_PORTABLE_CLASS(ID, TYPE) - for use with PortableObject
COH_REGISTER_POF_SERIALIZER(ID, CLASS, SERIALIZER) - for use with PofSerializer
Question 2: In an all C++ environment, is there any need to implement a Java version of the class? Would the Coherence server processes need to know about it in order to handle storing it?
I may have more questions later, thanks for the help!!
Edited by: user8723994 on Jan 25, 2010 6:44 AM

Answer 1

I did NOT manage to make the -Dtangosol.pof.enabled=true (or just -Dtangosol.pof.enabled) make any difference.
By changing the value in the XML-file (in the JAR) I did however manage to get the desired effect.
For an object with a few nested objects (all attributes in the objects involved were small ints, the class names were short) the results are:
Method / toBinary[s} / fromBinary[s] / size in bytes
Lite / 1.58e-6 / 7.82e-6 / 102
POF / 4.12e-6 / 2.82e-6 / 79
POF serialization is in this case like 250% SLOWER, de-serialization about as much FASTER and the size reduction is about 20%
Given that a read-heavy application do a lot more de-serialization the increase in de-serialization performance will easilly compensate for the slower serialization - looks promisiong!
I also noted that PowWriter/Reader methods for insts automatically seem to only write/read as many bytes as are required. If one would like to handle that yourself (ie use a single byte/word/int that holds the size info for several ints and then call the byte, word, int etc methods explicitly for each of them) is there some way to still do that?
/Magnus