Comments on Grid Designer's Blog: Oracle Coherence using POF, without a single line ...

Interesting idea. Current implementation is not de...

2010-07-14T08:32:10.629-07:00

Interesting idea. Current implementation is not designed to support POFExtractors. But it turns out to be quite simple to implement.
I need some more time to properly cover new ReflectionPofExtractor with test.
I will publish PofExtractor support soon.
Thanks for an idea.

how can you use this with POFExtractor and filters...

2010-07-13T06:03:14.646-07:00

how can you use this with POFExtractor and filters ?

I assume that was jmap, not lmap, not sure where h...

2010-01-06T17:42:20.194-08:00

I assume that was jmap, not lmap, not sure where have you find reference to lmap in this post.

Where can I find information about this "lmap" too...

2010-01-05T02:11:23.401-08:00

Where can I find information about this "lmap" tool you talked about? I know about "jhat" for heap analysis but not not lmap...

Googling ling did not find anything...

Hi ZeoS. I use this implementation with about 50 ...

2009-12-23T08:57:17.948-08:00

Hi ZeoS.

I use this implementation with about 50 classes (number of classes was a motivation to write generic tool in first place). Reflection in java can be extremely fast if you do it right. JVM generates code for relflection accessors, which is further optimized by JIT like your own code.
Creating an accessors is expensive operation, of cause, and any reflection based code should carefully cache generated accessors.
In my implementation a class serialization scheme is generated on first attempt to serialize/deserialize object of given class. Such serialization scheme stores all field accessors needed and they are generated only once for each class.
In other words, my implementation has own reflection cache to avoid lookup overheads.

Have you tried this with some real scenario? Refl...

2009-12-23T06:41:20.560-08:00

Have you tried this with some real scenario? Reflection will cache method lookups and the class loading will be done only in the first loop. by testing it with only one object/class you almost discard that overhead.

Thanks for a link Patrick, looks interesting. But ...

2009-10-19T06:09:23.915-07:00

Thanks for a link Patrick, looks interesting. But I really like to keep my domain objects clear from product specific annotations.
From the other side, ability to generate .NET and C++ would be great.
I will look forward for project evolution.

Nice post, thanks for sharing this! You might be ...

2009-10-07T08:34:27.158-07:00

Nice post, thanks for sharing this! You might be interested to look at this approach as well from Rob Varga:

http://coherence.politext.info/pof-serializer-generator-java-apt-plugin