[jboss-as7-dev] another approach to cli commands / operations

[Alexey Loubyansky]

Up until now I've been writing commands and handlers manually. It is 
justified in some cases like deploy for example. But I realized it's too 
much redundant work for commands to e.g. manage datasources, etc.

What I mean is I wrote add-data-source, remove-data-source, 
modify-data-source. Then there could be enable, disable, flush 
connections, etc. So instead, I wrote one command that I called data-source.

Now it looks like this (it's tab-completion)

[standalone at localhost:9999 /] data-source java:/H2DS

add                             disable                         enable 
                         flush-all-connection-in-pool
flush-idle-connection-in-pool   read-attribute 
read-children-names             read-children-resources
read-children-types             read-operation-description 
read-operation-names            read-resource
read-resource-description       remove 
test-connection-in-pool         write-attribute

So you can choose whatever operation you want, e.g.

data-source java:/H2DS flush-all-connection-in-pool

or whatever. If there are properties, it could look like

data-source java:/H2DS my-operation --prop1=val1 --prop2=val2

or to modify properties of the data source

data-source java:/H2DS --min-pool-size=10 --max-pool-size=100

At the end, this will save the total number of the commands, avoid 
possible inconsistency between commands and operations (like supported 
properties and even operations), the changes in operations and 
properties will be immediately exposed through the cli w/o 
changing/introducing new cli commands.

In addition, this data-source thing is actually no more than an alias to 
the node path, i.e. /subsystem=datasources/data-source. So, actually, 
this can be done for any component in a generic way, e.g. jms, etc. And 
even it could be customized by the user at runtime.


Again, that's not for everything. For things like deploy, we could keep 
hand-written command handlers but for the majority of the managed 
components this could work (although there are some operations that 
don't make much sense to expose, e.g. write-attribute or add as an 
operation on a specific data source but these could be filtered somehow).

So, if anybody has suggestions or objections, please, let me know. 
Otherwise, I intend to do it this way. Not for CR1 though.

Thanks,
Alexey