i have experienced jcs for hibernate before

i read the doc that "read-only" has more performance than"read-write"

but my data will be edited , may dur a long time

as my experience , if i update one piece of data , it's all relations must be "read-write" , otherwise , exception will welcome me

i wonder there is a way , i use "read-only" cache , and in runtime , i update the data then refresh the cache( maybe clear the cache) coded in my project?

can i handle this in "read-only" env?

Not true. Perhaps you should explain this problem a bit better. Is it because you enable cascade="save-update"?

ah ha , gavin always quick and kind answer :)

en , let me try to describ

as i said in another post(http://forum.hibernate.org/viewtopic.ph ... 66#2176666) , i have a ForumUser object

theForumUser 's cache is "read-write" , but one of it's relations Face as another object is "read-only" , the many-to-one schema is :



so , if i update forumuser like above , application will throw exception

but , if i change the cache of Pic to "read-write" , everything is ok

maybe something wrong in my schema

Yes, the reason for that is that you have cascade="all". Now, for a read-only class, cascade="all" makes absolutely no sense - it means that it is possible to delete this object of class. In your case, I think that what is happening is that the update is cascaded to the child, which is an error, because the child is read-only.

Now, this -does- raise an interesting point. It is not unreasonable to have



Because you want new instances of Bar to be saved by cascade, but because they are immutable, an update is unnecessary.

Currently, Hibernate would throw an exception in this case. This is suboptimal, I think. You should be able to persist immutable objects by cascade. What we need is for update() to degrade gracefully to lock() in the case of an immutable object.


Anyway, regarding your problem, try usage="nonstrict-read-write", if performance is bothering you - though I don't particularly see how it could be such a big difference.

thanks for your particular reply!

i exactly konw what i should do next.

the way , i use xdoclet to generate the mapping-file , but it doesn't support the "nonstrict-read-write" property

maybe we should advice xdoclet to update their product :)