[infinispan-dev] Retrieval operations with the IGNORE_RETURN_VALUES flag

[Galder Zamarreño]

On Jun 13, 2013, at 2:40 PM, Dan Berindei <dan.berindei at gmail.com> wrote:

> 
> On Wed, Jun 12, 2013 at 3:39 PM, Mircea Markus <mmarkus at redhat.com> wrote:
> 
> On 12 Jun 2013, at 13:14, Galder Zamarreño <galder at redhat.com> wrote:
> 
> >
> >
> > On Jun 10, 2013, at 12:01 PM, Adrian Nistor <anistor at redhat.com> wrote:
> >
> >> Maybe we could just clarify the javadoc of IGNORE_RETURN_VALUES and say that it only applies to write operations and is ignored for everything else? Why punish the user with an exception when doing a 'get'?
> >>
> >> We already document there's a (very common-sense) exception for conditional writes were the flag is ignored (ISPN-3141).
> >
> > I wonder if anyone noticed my reply earlier...
> >
> > "The flag business does need a big re-think. Not only to separate internal from external flags (we have a jira for that [1]), but also to have a way to define which flags can be passed to a particular operation, in a way that's type-safe, and without resulting in a runtime error of the likes of "X flag cannot be used with Y operation". IOW, any error on which flag can be used with what operation should ideally be caught at compilation time. I don't have specific ideas on this right now, but I think it'd be good to achieve this."
> >
> > IOW, I suggest we leave it as it is. We need to re-think it anyway. So let's tackle it in 6.0 so that a get operation can never be passed IGNORE_RETURN_VALUES flag, and this being something that's caught at **compilation time**.
> 
> this would be the elegant way of doing it.
> 
> 
> An even more elegant solution would be to make put return void by default, and force the user to state explicitly that he wants a return value - a la JSR-107. I know that would break backwards compatibility, so it may not be an option even in 6.0, but still I think we should focus on that.
> 
> Maybe it's worth making other flags "type-safe", but I don't think it's worth doing it for IGNORE_RETURN_VALUES.

Fair point. I hope once JCache (JSR-107) is final we can move towards making their Cache API the main entrance point and reduce our Cache to having only those methods not present in JCache.

> 
>  
> >
> > I'm just about to add another internal flag to Flag as a result of the JCache 0.7 upgrade…, so need to tackle ISPN-2201 to avoid causing more confusion, and alongside avoid the issues that have been highlighted WRT which operations are allowed which flags. I'm happy to do this for 6.0.
> >
> > [1] https://issues.jboss.org/browse/ISPN-2201
> I've update the JIRA to track the fact that IGNORE_RETURN_VALUES + get should not be possible.
> 
> Cheers,
> --
> Mircea Markus
> Infinispan lead (www.infinispan.org)
> 
> 
> 
> 
> 
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
> 
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev


--
Galder Zamarreño
galder at redhat.com
twitter.com/galderz

Project Lead, Escalante
http://escalante.io

Engineer, Infinispan
http://infinispan.org