We already have a method GridCacheProcessor.checkCache() which accepts two cache configurations - local and remote - and decides whether they are compatible. Looks like we just need to refactor it a bit so that it can be used for "getOrCreateCache" scenario.
However, whatever checks we perform, there still be some scenarios which is critical for users, but cannot be checked by Ignite directly. First of all I am talking about user-defined resources. And it will end up in the same usability problem. This is why I think we *MUST* give user ability to know whether cache was actually created by his call to getOrCreateCache() or not. May be using additional returned boolean flag, may be using optional closure which is executed if cache is created, whatever. On Fri, Aug 14, 2015 at 9:39 PM, Dmitriy Setrakyan <dsetrak...@apache.org> wrote: > On Fri, Aug 14, 2015 at 12:47 AM, Valentin Kulichenko < > valentin.kuliche...@gmail.com> wrote: > > > I think we should just throw an exception in this case. Providing two > > configurations for one cache looks like incorrect usage for me. > > > > I don't think we should make assumptions here. Let's just throw an > exception if configurations do not match in a critical way. > > > > > > -Val > > > > On Thu, Aug 13, 2015 at 11:43 PM, Dmitriy Setrakyan < > dsetrak...@apache.org > > > > > wrote: > > > > > On Thu, Aug 13, 2015 at 5:56 PM, Alexey Goncharuk < > > > alexey.goncha...@gmail.com> wrote: > > > > > > > In this case, since user called getOrCreateCache(), Ignite did not > > > create a > > > > cache because it already existed and also did not throw an exception > > > > because method should return existing cache if a cache with the same > > name > > > > already exists. > > > > > > > > We may want to implement a check to verify that cache configurations > > > match, > > > > but I have no idea how to compare cache configurations given that > they > > > may > > > > contain user-defined objects, such as store factories. > > > > > > > > > > Can we to the least compare what we can and skip whatever we cannot? > For > > > example, for factories, we can make sure that the class names match. > > > > > > I definitely think that we need some level of validation here. Users > keep > > > getting burned on this issue. > > > > > > Thoughts? > > > > > > > > > > > > > > 2015-08-12 9:45 GMT-07:00 Dmitriy Setrakyan <dsetrak...@apache.org>: > > > > > > > > > Igniters, > > > > > > > > > > After looking at the comment in the IGNITE-1240 [1], I got confused > > on > > > > how > > > > > Ignite configuration works. It seems to me that we have a big > > usability > > > > > issue there. > > > > > > > > > > For example, can someone tell me what happens now if a user > provides > > > the > > > > > configuration in an XML file and then also specifies the > > configuration > > > in > > > > > code? > > > > > > > > > > [1] https://issues.apache.org/jira/browse/IGNITE-1240 > > > > > > > > > > Thanks, > > > > > D. > > > > > > > > > > > > > > >
