I currently think that declarative ta-demarcation at the service-facade in a typical web-app with hibernate is just plain wrong. Because it pretends to hide something (the hibernate session and its lifecycle) which it doesn't. Example from a Spring/Hibernate project I was involved in some time ago: a developer called two service-methods, both marked as "requires ta" in application.xml, in one http request. He thought, they would run in the same ta, which, of course, they didn't. This was a rather nasty bug. Since then I have ta control in some point of the web-layer (DirectService.trigger). This works very well for me. Of course this is also the approach taken in honeycomb.
See also: http://marcus-schulte.blogspot.com/2005/04/if-you-cant-hide-it-below-hide-it .html > -----Original Message----- > From: Mark Reynolds [mailto:[EMAIL PROTECTED] > Sent: Thursday, April 13, 2006 4:51 AM > To: tapestry-user@jakarta.apache.org > Subject: Re: tapestry-hibernate integration problem > > > The problem I experienced using hivetranse and the "open session in > view" was this: > > 1) execute one transactional method, transaction is committed > when the > method returns and the hibernate session is kept open > 2) execute another transactional method, hivetranse opens a second > hibernate session, transaction is committed when the method > returns and > the hibernate session is kept open > > Now I have two hibernate sessions for the same request and can easily > get hibernate errors regarding the same persistent object > being on two > different sessions. > > It seems like hivetranse should be able to reuse the same session > throughout a single request, as long as it doesn't hit a > nested method > demarcated as "requires new". > > Has anyone found a solution for this? Or at least seen the same issue? > > -- Mark R > > Nick Faiz wrote: > > Just a note that you dont need to write code to control the > Hibernate > > session & transaction lifecycles if you use HiveUtils (aka > Hivetranse). > > > > http://sourceforge.net/projects/hivetranse/ > > > > In my current project, I configured the hibernate session > as a hivemind > > component, which is injected into whichever component I > like. I also use > > the HiveUtils transaction interceptor, which transparently handles > > transactions for all components using it, following the > 'open session in > > view' pattern. > > > > > http://hivetranse.sourceforge.net/web/quickstart.html#start.co > nfig.hibernate > > > > > > Nick > > > > On 12/04/2006, at 10:21 PM, bėgantis debesis wrote: > > > >> Hi everyone, > >> > >> I extended a BasePage and added there a getSession method, > which opens a > >> hibernate session and starts a transaction (if it is not > yet opened). > >> Also, > >> I commit the transaction and close the session on the overrided > >> BasePage.detach() method. > >> > >> The problem is that when I save something on page 1 and > then redirect to > >> page 2, page 2 gets the old information from the database. That is > >> because > >> detach method on page1 (where the transaction commit > resides) is invoked > >> after page2 getInformation() method called. > >> > >> So, the question is, is there a method like detach but > which is invoked > >> before another page starts its activities? Or maybe using > transaction per > >> request is bad at all? > >> > >> Or maybe you know some other easy way to intergrate tapestry and > >> hibernate? > >> > >> Thank you in advance, > >> Valdemaras > > > > Nick Faiz, > > Developer > > www.q9software.com > > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: > [EMAIL PROTECTED] > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
