Hi Jeff, Thanks for your response!
I'm going to take a while to absorb (1) and (2) but here's a quick response for (3) and (4). (3) > It looks to me like you are seeing the same objects > over and over again rather than seeing a huge tree Yes, absolutely. I'd noticed that that was the case. And I assume that the fact that they are the same objects referenced over and over limits the damage. Instead of huge numbers of objects we just have huge numbers of references? Or maybe we don't - I guess that we're just seeing the same references over and over again too. But something is slowing the app down big-time. I've pared the number of records in the DB down to hundreds per table so that it loads in a few seconds, but with thousands of records and non-lazy hierarchical-data-approach it was taking over 30 seconds to load... (4) Can you give me a few details on how to turn on these two types of debugging? Even just file names would help - then I can Google/research the details - and ask again if I get stuck. :-) Thanks, Douglas Jeff Vroom wrote: > > Sorry for not chiming in sooner – I was on vacation the last couple of > days. > > I’m afraid I haven’t had a chance to look at the code you posted but I > have a few comments that will hopefully help. > > 1) The HibernateAssembler does have some logic to deal with > hibernate’s generated proxy classes when you are using FDMS with > hibernate. This involves two parts: 1) pre-fetching any lazily loaded > data that we will need to serialize to the client and 2) unwrapping > the proxied classes so that the class names are properly mapped if you > are using strongly typed AS classes to map to your hibernate domain > model classes. We need to do 1) so that hibernate has loaded all of > the object’s state that the client requires while the transaction is > still open. This also allows us to unwrap the proxies during > deserialization since we know all of the state we need will have been > fetched in the wrapped implementation instance. There is a performance > issue with the code that implements 1) in that currently hibernate’s > lazy loading and FDMS’s lazy loading are a little different for > multi-valued properties and this could be contributing to your > performance problem. Basically, we FDMS needs to prefetch the value of > the multi-valued association property even when the property is marked > with lazy=”true” in your FDMS configuration. This is because FDMS > wants to send the ids of the referenced objects down with the state of > the parent object… hibernate initially does not fetch any values in > the multi-valued association property and when you first retrieve the > collection it gets all of the data in the collection. We have an open > enhancement request to support a model for FDMS which matches > hibernate’s to make this more efficient. > > 2) We should support bi-directional references in your object graph > though it is a good practice to ensure all “back ptrs” are marked with > lazy=”true” in the FDMS configuration. > > 3) In the flex-builder variables view you sent in the screen-shot, > keep in mind that flexbuilder is not good at letting you know when you > have hit a recursive ptr in the variables view. It looks to me like > you are seeing the same objects over and over again rather than seeing > a huge tree that is being sent to the client (you can match up the > addresses flexbuilder displays for each object to detect this yourself). > > 4) If you turn on debug logging in the server for the “Endpoint.*” > logging category, you’ll see the object graph which is sent to the > client after your fill. I am curious if the time is being spent > loading the data from the database or sending data to the client, or a > combination of both. If you can also turn on debug logging of the SQL > that hibernate is using that would help figure out if we are just > fetching too many objects or what. > > I’ll hopefully get a chance to look into the test case in more detail > later. Regards, > > Jeff > > ------------------------------------------------------------------------ > > *From:* [email protected] [mailto:[EMAIL PROTECTED] > *On Behalf Of *Douglas McCarroll > *Sent:* Tuesday, November 28, 2006 3:19 PM > *To:* [email protected] > *Subject:* Re: [flexcoders] many-to-many managed association in > Hibernate destination > > Hi Richard, > > > I was under the impression that you have to use DTOs, rather > > than mapping your client classes to your domain objects. > > I Google "flex hibernate lazy" and I find this: > > http://livedocs.macromedia.com/flex/2/docs/wwhelp/wwhimpl/common/html/wwhelp.htm?context=LiveDocs_Parts&file=00001201.html > > <http://livedocs.macromedia.com/flex/2/docs/wwhelp/wwhimpl/common/html/wwhelp.htm?context=LiveDocs_Parts&file=00001201.html> > > It includes config code that looks like this: > > <destination id="HibernatePerson" channels="rtmp-ac"> > <adapter ref="java-adapter" /> > <properties> > <source>flex.data.assemblers.HibernateAssembler</source> > <scope>application</scope> > <metadata> > <identity property="id"/> > <many-to-one property="group" > destination="HibernateGroup" lazy="true"/> > </metadata> > > This looks like they're mapping directly to a domain object, yes? > > Douglas > > Richard Rodseth wrote: > > > > I'm not currently using Flex, but I don't believe it's possible to map > > ActionScript classes to lazily-loaded Hibernate classes because of all > > the hibernate proxy stuff. I was under the impression that you have to > > use DTOs, rather than mapping your client classes to your domain > > objects. Am I wrong? > > > > - Richard > > > > On 11/28/06, *Douglas McCarroll* > > <[EMAIL PROTECTED] > <mailto:org.yahoo_primary.001%40douglasmccarroll.com> > > <mailto:[EMAIL PROTECTED] > <mailto:org.yahoo_primary.001%40douglasmccarroll.com>>> wrote: > > > > Hi All, > > > > This is essentially a repost of my "How to implement Lazy Hibernate?" > > question. > > > > I'm still completely stuck. Here's the error message I'm getting: > > > > [RPC Fault faultString="Page request made for item with id = > > '{filmId=null}'. The item was not in the cache and the adapter's get > > method failed to return the item for destination = film.hibernate > > that > > is expected to be in the requested page." > > faultCode="Server.Processing" > > faultDetail="null"] > > > > Here's the big picture: > > > > I'm attempting to create a Cairngorm/FDMS/Hibernate/MySQL example > > program. > > > > I have three tables. 'actor' and 'film' have a many-to-many > > relationship > > so they're connected by a link table 'film_actor'. > > > > I've used HibernateTools to create .java and .hbm.xml classes by > > reverse > > engineering the DB. I assume the HibernateTools knows what it is > > doing > > so I'm assuming that the generated code is correct. > > > > And I've successfully implemented .as classes and > > data-management-config.xml and displayed data. I can click on an > > actor > > and display the films that s/he's in. This all works through a single > > destination that uses my Actor.java class. This uses the > > 'hierarchical > > values' approach to managing hierarchical collections, i.e. just > > using > > the 'actor' destination. > > > > This approach is problematic because there are circular references > > between Actor and Film objects. It seems to me that the obvious > > solution > > is lazy loading. I don't see anything in the docs explaining how > > to do > > this with the hierarchical values approach, so I'm attempting to > > use a > > 'lazy' managed associations approach. But I get the above error. > > > > Someone who knows FDMS might be able to look at my code and fairly > > easily point me in the right direction. > > > > I've zipped up the entire project - plus a .sql file you can use to > > create my MySQL DB - and put them up for download: > > > > http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/Bw_005.zip > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/Bw_005.zip> > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/Bw_005.zip > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/Bw_005.zip>> > > > > Or, you can just look at the code here: > > > > http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/code.txt > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/code.txt> > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/code.txt > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/code.txt>> > > > > The DB's schema is here: > > > > http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/index.html > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/index.html> > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/index.html > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/index.html>> > > > > And here's a lovely variables view of the circular references I'm > > trying > > to avoid :-) : > > > > http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/deeply_nested_actors_and_films.jpg > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/deeply_nested_actors_and_films.jpg> > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/deeply_nested_actors_and_films.jpg > > > <http://www.brightworks.com/technology/tech_questions/hibernate_lazy_associations/deeply_nested_actors_and_films.jpg>> > > > > Help? > > > > Douglas > > > > > > > > -- Flexcoders Mailing List FAQ: http://groups.yahoo.com/group/flexcoders/files/flexcodersFAQ.txt Search Archives: http://www.mail-archive.com/flexcoders%40yahoogroups.com Yahoo! Groups Links <*> To visit your group on the web, go to: http://groups.yahoo.com/group/flexcoders/ <*> Your email settings: Individual Email | Traditional <*> To change settings online go to: http://groups.yahoo.com/group/flexcoders/join (Yahoo! ID required) <*> To change settings via email: mailto:[EMAIL PROTECTED] mailto:[EMAIL PROTECTED] <*> To unsubscribe from this group, send an email to: [EMAIL PROTECTED] <*> Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
