This one time, at band camp, Steffen Fiedler said:
SF>Hi,
SF>
SF>I have an application that works with objects, that are linked in many
SF>ways among each other.
SF>What is the best way to save them?
SF>I try to to it by long transactions, because the Objects are used while
SF>the whole application
SF>and and i don't want to hold one transaction all the time to hold them
SF>persistent.
SF>Its a multi-user web-application and i can't be sure, that the conection
SF>will be closed correctly.
SF>So i just open a connection, when some objects have to be
SF>read/saved/updated and want to
SF>save them by long transactions. When i do this:
SF>
SF>db.setAutoStore(false);
SF>db.begin();
SF>db.update(myObject);
SF>db.commit();
SF>
SF>i get the error:
SF>
SF>Commit failed: org.exolab.castor.jdo.TransactionAbortedException: Nested
SF>error:
SF>org.exolab.castor.jdo.PersistenceException: Object,
SF>com.sourcepark.ticketsystem.Ticket@305236,
SF>links to another object, com.sourcepark.ticketsystem.Priority@6a0105
SF>that is not
SF>loaded/updated/created in this transaction
SF>
SF>If i set db.setAutoStore(true) i get the error:
SF>
SF>Error while saving object:
SF>org.exolab.castor.jdo.DuplicateIdentityException: update object
SF>which is already in the transaction
SF>
SF>because i use some objects in many ways, i.e. the object 'state' can be
SF>a pre-state and a
SF>successor-state for another object and there are many other relations
SF>like this. I've tried
SF>to set the affected objects to access="read-only" in mapping.xml but
SF>without success.
SF>
SF>In the JDO-FAQ is written, that castor only supports bi-directional
SF>relationships, does
SF>that mean, if object1 contains a reference to object2, object2 should
SF>contain a vector
SF>of object1's?
SF>
SF>I expanded my mapping.xml, but now castor works very slowly, when i load
SF>a single
SF>user, the whole datamodel is loaded from DB cause all objects are linked
SF>among each other.
SF>Is it usefull to use castor for a web-application with complex
SF>datamodel? Whats about
SF>performance when there are many users and large databases. Could
SF>lazy-loading solve this
SF>problem?
SF>
SF>I've appended the whole mapping.xml for getting an impression about the
SF>object-structute.
SF>My main question is, if it's advisable to use castor for a multi-user
SF>web-application with
SF>such a datamodel and if yes, what is the most performance-friendly
SF>approach to save and
SF>load the objects.
Steffen,
I've found that Castor has it's limits with regards to very, very complex
object models. Much of my experience with this has been learned at my
day job where I work on a GIS application using highly complex data
in a highly complex object model (polygons, etc). BTW, I would NOT
consider your object model to be highly complex. Castor should handle
this just fine.
I've also found that Castor does not enforce the bi-directional relations
at all times (as is stated in the JDO FAQ). This being said, I've used
Castor in the past with an object model similar to yours by not mapping
things bi-directionally. (This took some time to figure out because there
are situations where Castor wants a bi-directional mapping and other
situations where it doesn't matter (there are threads on the mailing
list discussing this, unfortunately I don't know the ins and outs of
this off the top of my head).) The situation to which I'm referring
was a mutli-user web app using long transactions. In this setting,
Castor performed just fine.
Unfortunately I have no easy answer for this situation. However,
if you take the time to experiment with the removal of many of the
bi-directional references in your object model, I believe you'll start
to get the results you want.
Bruce
--
perl -e 'print unpack("u30","<0G)U8V4\@4VYY9&5R\"F9E<G)E=\$\!F<FEI+F-O;0\`\`");'
-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
[EMAIL PROTECTED] with a subject of:
unsubscribe castor-dev
