Well this is just one pre-requisite for "cleaning up" how components are modeled. Specifically, the piece I want to clean up is the fact that components are currently handled differently than any other mapping construct. When components are being parsed and bound (during config time) they require construction of and access to things that are typically only available after SF construction (runtime) for all other mapping contructs. This mis-alignment causes some goofiness in the way ComponentTypes are built and, even more importantly, some goofiness in some of the things upon which ComponentTypes currently depend.
Yes, one of these is how bytecode providers are handled. Another is PropertyAccessors. Yet another is ComponentTuplizers. -----Original Message----- From: Max Andersen Sent: Monday, July 24, 2006 2:18 AM To: Steve Ebersole; Hibernate development Subject: Re: [Hibernate] Roadmap - components > So that leaves the option of having Types be aware of the session > factory. Which upon further reflection is not as bad as it sounds, > because really there are only a few Type implementations that even need > access to the session factory at all in any of these methods > (sqlTypes(), etc exluded). These would be limited to mainly the > EntityType subclasses, the ComponentType subclasses, and the > CollectionType subclasses. The essential point being that none of the > "primitive"/"simple" types are in this category... Sounds fair. > The one wrinkle in this approach is the various type factory methods on > the o.h.Hibernate class. These would need to change signature, or go > away. I guess we could limit these changes to be only for those who really needed or would it be better to just get it over with. Still, *alot* of code will break for this :( (luckily it is compile-detectable) Is it worth doing this for sessionfactory scoped types ? Do we get more out of this than sf-controllable bytecode provider ? (I guess some of the issues regarding overriding the default types would also get in here) Could we somehow allow both possibilities for the sake of compability ? And I know Christian will "love" us for changing this so close to the book release ;) /max > -----Original Message----- > From: Max Andersen > Sent: Monday, July 17, 2006 10:51 AM > To: Steve Ebersole; Hibernate development > Subject: Re: [Hibernate] Roadmap - components > > point taken. > >> Well Type and UserType do not necessarily need to be in synch in this >> particular regard. We could conceivably change Type and then later >> (i.e. as part of a major release) change the UserType API to align it. >> After all the whole point of the UserType stuff was to insulate the > user >> from changes in the underlying Type system... >> >> -----Original Message----- >> From: Max Andersen >> Sent: Monday, July 17, 2006 10:43 AM >> To: Steve Ebersole; Hibernate development >> Subject: Re: [Hibernate] Roadmap - components >> >> On Mon, 17 Jul 2006 17:41:25 +0200, Steve Ebersole >> <[EMAIL PROTECTED]> wrote: >> >>> Type is *NOT* a public API... >> >> but UserType is - don't they need access to this info too ? >> >> /max >> >>> >>> -----Original Message----- >>> From: Max Andersen >>> Sent: Monday, July 17, 2006 10:38 AM >>> To: Steve Ebersole; Hibernate development >>> Subject: Re: [Hibernate] Roadmap - components >>> >>> ...but requires changes to public API so probably best suited for > 3.3. >>> >>>> Regarding the component related changes mentioned in the previous >>>> email... >>>> >>>> As I mentioned a lot of the pre-requisite work has already been >>>> performed on HEAD. I also took the opportunity to refactor the >>>> packaging of the org.hibernate.tuple package. Specifically, most of >>> the >>>> pre-requisite work was the introduction of the >>>> o.h.t.component.ComponentMetamodel class. Currently, ComponentType >>> just >>>> uses this new class directly. >>>> >>>> What needs to happen next, then, is for the introduction of a >>>> org.hibernate.persister.component.ComponentPersister which is > managed >>> as >>>> part of the session factory much like the other persisters. >>>> ComponentType will then need to look up its corresponding >>>> ComponentPersister based on a "role name" and use the capabilities > of >>>> that persister. The pattern here is very similar to >>>> EntityType/EntityPersister. The difficulty I ran into though was >> that >>>> ComponentType would then require access to the session factory (in >>> order >>>> to locate the persister) from within methods where it is currently >> not >>>> passed a reference to the session factory (specifically, this was >>>> methods like isSame(), isEqual(), compare(), getHashCode(), etc). >>> This >>>> gets to more general discussions we have had in the past regarding >> the >>>> scoping of Types. The solution is one of two things: >>>> 1) Devise some sort of scoping scheme where Types can unequivocally >> be >>>> "bound" to a session factory. This is obviously difficult given the >>>> current Hibernate.LONG, Hibernate.STRING, etc static references. > One >>>> thought here would be splitting types (and their interface >>>> appropriately) to define "static" Types and "scoped" Types... >>>> 2) Modify the Type interface to accept either a session or a session >>>> factory/entity mode combo for most methods (would not really matter >>> for >>>> methods like sqlTypes(), etc) >>>> >>>> As I mentioned before this then allows us to make the >>>> 'hibernate.bytecode.provider' and >>>> 'hibernate.bytecode.use_reflection_optimizer'. Down the road, it >> also >>>> allows us to implement discrimination-based inheritance for >>> components. >>>> >>>> >>>> >>> >> > ------------------------------------------------------------------------ >>> - >>>> Using Tomcat but need to do more? Need to support web services, >>> security? >>>> Get stuff done quickly with pre-integrated technology to make your >> job >>> >>>> easier >>>> Download IBM WebSphere Application Server v.1.0.1 based on Apache >>>> Geronimo >>>> >>> >> > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 >>>> _______________________________________________ >>>> hibernate-devel mailing list >>>> hibernate-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/hibernate-devel >>> >>> >>> >> >> >> > > > -- -- Max Rydahl Andersen callto://max.rydahl.andersen Hibernate [EMAIL PROTECTED] http://hibernate.org JBoss Inc [EMAIL PROTECTED] ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ hibernate-devel mailing list hibernate-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/hibernate-devel
