Hi all,
I have a question about maintaining object references in the
JavaSerializer class (see snippet below).
This seems to maintain a reference to every object it serializes.
In a streaming scenario where you want to stream mass data, this is
causing memory consumption issues.
To work around this I suclassed JavaSerializer and removed the
reference maintenance with the writeObject() method.
I should emphasize that this is only a workaround, as:
- I had to copy a lot of the fields and code form the super class
because most of the stuff I needed isn't publicly visible.
- This ignores the reference maintenance code, which may be
functionally wrong and have other adverse effects.
I was wondering if there is any scope to change the API of the
JavaSerializer to cater for this issue?
Instead of completely bypassing the reference maintenance, maybe one
could set a limit of the size of the hash map?
Thanks,
Ben
-----
public void writeObject(Object obj, AbstractHessianOutput out)
throws IOException
{
// ignore reference maintainance code .....
if (out.addRef(obj))
return;
Class cl = obj.getClass();
try {
if (_writeReplace != null) {
Object repl = _writeReplace.invoke(obj, new Object[0]);
out.removeRef(obj);
out.writeObject(repl);
out.replaceRef(repl, obj);
return;
}
} catch (Exception e) {
}
int ref = out.writeObjectBegin(cl.getName());
if (ref < 0) {
writeObject10(obj, out);
}
else {
if (ref == 0)
writeDefinition20(out);
writeInstance(obj, out);
}
}
_______________________________________________
hessian-interest mailing list
[email protected]
http://maillist.caucho.com/mailman/listinfo/hessian-interest
