Re: [Openjmx-devel] [Fwd: Source code generation of MBean interfaces and descriptions]

[B. <jer...@xt...>]

Bordet, Simone wrote:

>>I have attached to the following mail the MBean implementation with
>>proper XDoclet comments.
>>I have also attached the results of XDoclet on the 
>>implementation source.
>>
>
>Yes. But I have few questions:
>1) there is a managed attribute "dummy", but no setDummy and getDummy in the MBean interface. Beware also of the fact that I can call the data member "m_dummy", does this impact the MBean interface (I don't want getm_dummy() ;) ? And also, beware of the fact that a "dummy" attribute *must* generate a getdummy() and setdummy(), not getDummy() and setDummy(). Can I specify if it's read-only ?
>
Ok. There are plenty of questions out there :-)
First, as there is a diffence between a java field and a JMX attribute, 
where should we put the description?
I think it should not be on the field itself because I am not sure it 
will be easy to check for the presence of the getter/setter methods.
But then, should we put it on the getter or setter method? both 
(redundant information and possible mistakes -- don't like that)?
Or we could put in the class javadoc and indicate something like that: 
"@openjmx:attribute name="Dummy" description="A dummy attribute". It 
would be up to the developer to implement the getDummy and setDummy 
methods. If the variable itself is m_dummy, then we don't care: the 
developer has to code the right stuff in the getter/setter methods.

Regarding access-mode, i suggest we add another attribute on the 
"openjmx:attribute" tag such "access" or "access-mode" and the possible 
values would be "ro", "wo" or "both".

>2) just for consistency with JMX, i would change "managed-method" to "managed-operation"
>
Fixed.

>3) setStatus is an attribute, not an operation.
>
As explained in 1) it would be up to the developer to specify that. If 
the "openjmx:attribute" tag is missing, the output will be wrong.

> Is there a way to enforce check of this kind of mistakes ? If a method is described as operation, but it's an attribute, can i know this when I'm generating the code with XDoclet ?
>
This is linked to question 1). I will look for a way to check those 
things but I am not sure it will be so easy.

>4) I see that the parameter information for operations is not included, it's just because start() has no parameters ?
>
No. I have to apology, I have forgotten some few methods from the 
MBeanDescription class. Was playing too much with the example given in 
the documentation and forgot to check the definition of the interface :-(

>Well, it's a great job, Jerome !
>Let me know if you need clarification on the questions above, but I guess the xdoclet generation will rock !
>I didn't know that XDoclet became so powerful.
>
>>I had to patch XDoclet because the template engine does not 
>>support yet
>>constructors. I am working with one of the XDoclet administrator in
>>order to put my modifications into CVS.
>>If you want to use it for now on, send me a mail and I will 
>>send you the
>>modified xdoclet.jar file.
>>
>
>Ok !
>
The patch has been sent and the administrator is currently reviewing it. 
Should be made available soon I hope...

>>If we want to add this features to OpenJMX (I think this is a really
>>nice thing), 
>>
>
>Absolutely agree !
>
We're at least two then ;-p

>>let me know where in the distribution I should do so...
>>
>
>Ok, let me play a bit with CVS, I will tell you soon.
>
Ok. Anyway I have to fix/enhance the generator, but I wanted to have 
some initial feedback before :-)

>If you have time, can you write a document on this beautiful feauture, when will be ready ?
>If you don't have time just tell, we must absolutely write something about for users.
>
I guess I could, but english is not my native language, and I am not 
sure to be the best person for writing documentation: my english style 
is quite poor. I can however write a draft that someone could "re-style" 
i guess :-)


Jerome.