On 11 Oct 2010, at 13:44, Mircea Markus wrote:
PS: There is no JIRA for this. If we like this approach and it works, I suggest we create a JIRA and implement it for 4.2. The impl should be simple once we resolve the outstanding bits.
there is one actually:
https://jira.jboss.org/browse/ISPN-256I'll attache the patches to it.
Just to conclude, I've POC both approaches GenericCommand and ExtendingCommand, here are my thoughts on it:
- extending visitor
PROS: ExtendingCommand is a better OO API, more strongly typed
CONS: it is harder for the user to understand; integrating custom serialization might be difficult(impossible at the time), also it will have to manage command uniques.
Why is this hard for developers to understand? Remember, these aren't *application* developers, but people *extending* Infinispan. They already ought to have an in-depth knowledge of Infinispan.
What do you mean by "command uniques"?
- GenericCommand
CONS:
- less OO API
- would require a "convention" for not "overusing" it. Overuse would be in client code though, not in ours
Client code may well be our code. E.g., more modules that we ship. -1 to a weakly typed API like this. Conventions cannot be enforced, and IMO if something can be abused, then it will be abused. :)
PROS:
- easier for the user to understand and use
- no need for extra classes needed
- no understanding of how the interceptor chain internals work needed (for writing acceptVisitor).
Why would someone write a visitable command and not care about an interceptor that visits it? Surely you would only write a visitable command if you *want* to visit it with an interceptor.
- no understanding of how serialization works is needed. Nor command uniques.
? What's so hard about this? "implements Externabilzable" is the easy way, @Marshallable() a slightly more roundabout way for greater efficiency in marshalling.
- easy streight-forward solutions
I would go for GenericCommand, but Extending approach works for me as well.