10:06 a.m.
On Sep 10, 2009, at 4:59 PM, Galder Zamarreno wrote:
On 09/10/2009 11:38 AM, Vladimir Blagojevic wrote:
> Hi Galder,
>
> Although I agree with you that reflection is little bit harder to
> understand the advantage of this visitor pattern is many fold. Say,
> down
> the road, we have multiple of these visitors written, some of them,
> possibly, even written by third party clients.
Hmmm, can you be more specific? We're talking about here in the
scope of
Configuration validation.
> Our entire configuration
> tree class hierarchy is statically or compile time tied tied to these
> visitors and we don't want to go down this path. I'll give it another
> look, we can go static but we need to be aware of the trade offs.
We've implemented the visitor pattern statically for invocation
execution (see org.infinispan.commands.Visitor. Mircea, you're the
original author so correct me if I'm wrong) and I don't we had any
problems dealing with extending it. In fact, I had to extend it myself
to cope with keySet and entrySet commands and wasn't that bigger deal
(https://jira.jboss.org/jira/browse/ISPN-94?page=com.atlassian.jira.plugin...
)
The reflection approach is both good (for the point Vladimir outlined)
and bad: as code won't fail to compile when we change it (not strongly
typed) the clients won't even be informed that their visitor's code
code is not called/visit method are skipped - this is when we change
our code. This might result in more time to investigate on their side,
and more frustration for them. On the other hand the strongly typed
code would fail-fast, telling the users that something needs to change
- hence my vote for the strongly typed visitors.
>
> If I understand you correctly the approach you are suggesting is not
> modular. It is generally considered not to be a good practice to tie
> logical processing to data structures themselves when the
> likelihood of
> extending logical processing is significant. The whole reason behind
> adopting visitor pattern on our configuration tree is the ability
> to add
> new operations to existing configuration object structures without
> modifying these structures. By using visitor pattern and adhering
> to the
> so called open/closed principle we allow room to do any kind of
> operations on our configuration object structures.
I see your point of not tying up the processing and the data
structure.
> It might seems that
> validation is the last thing we are going to do - but it is not, for
> sure. And tomorrow when there is yet another operation we need to
> complete on configuration tree the implementation will be a breeze.
Hmmm, we could implement a swiss knife so that it can solve all the
problems in this world but before doing that, I'd like to hear more
specific examples ;). I mean, we could have a created a super generic
visitor pattern to deal with configuration validation, with
invocations...etc, but I'm against that since it's hard to read,
hard to
follow and hard to debug. For reflection related nightmares, simply
look
at JBC 1.x code.
So, to sum up, I see the point of using the visitor pattern but let's
time things a little bit. We're dealing with configuration validation
here, so let's stick that for the moment. Also, let's go for a
statically typed one similar to the one already used for dealing with
invocations. I'd imagine that applied to this use case, this would be
like: visitSingletonStoreConfig, visitTransportType...etc.
org.infinispan.config package:
ValidationVisitor {
void visitSingletonStoreConfig(SingletonStoreConfig ssc);
void visitTransportType(TransportType);
....
// instead of traversalCompleted, use more meaningfull name
validate()
}
SingletonStoreConfig:
public Object accept(Visitor visitor) throws Throwable {
return visitor.visitSingletonStoreConfig(this);
}
>
> Cheers,
> Vladimir
>
> On 09-09-10 4:11 AM, Galder Zamarreno wrote:
>> Hi Vladimir,
>>
>> I had a look at the implementation and I'm not sure I understand the
>> need for the reflection visit calls in
>> AbstractConfigurationBeanVisitor.
>> Using reflection makes harder to follow code and it's slower than
>> typed
>> calls and I'm not sure of the reason to use it here.
>>
>> Also, I don't see the need for a standard
>> ConfigurationValidatingVisitor
>> that does such validation. Instead, the way I see it working is
>> SingletonStoreConfig having some kind of callback method being
>> called,
>> i.e. the traversalCompleted() call and within it,
>> SingletonStoreConfig
>> can, using the ComponentRegistry, retrieve the Transport component
>> and
>> see if it's set or not. Or alternatively, SingletonStoreConfig
>> could use
>> the passed InfinispanConfiguration to do its validation.
>>
>> To sum up, I think each AbstractNamedCacheConfigurationBean
>> implementation should, if it requires to, have the ability to
>> validate
>> the configuration via some kind of callback. Such callback should
>> probably is possibliy traversalCompleted() itself.
>>
>> WDYT?
>>
>>
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
--
Galder Zamarreño
Sr. Software Engineer
Infinispan, JBoss Cache
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev