Well, I've just started working on this myself. But I do have an implementation as a Proof of Concept. There are some rules coded into it that are specific to our architecture, and so far I have only used it against a very simple object graph. My implementation really just needs to show what the differences are (no requirement to generate a 'patch'), but diff'ing an object graph is really nothing more than traversing the graph and diff'ing each object. My differ at present just looks at simple properties; support for diff'ing contained objects (whether as a simple reference or as a collection) comes from a separate class that is responsible for traversing the graph.
I'm all in favor of taking it from State.USEFUL_FOR_THIS_PROJECT to State.GENERICLY_COOL ! Dave "Ola Berg" <ola.berg@ports. To: [EMAIL PROTECTED] se> cc: Subject: Re: [BeanUtils] Javabean diff? 01/31/2003 07:06 AM Please respond to "Jakarta Commons Developers List" My BeanDiffer as present don't have any support for neither graphs or structural changes (like reorders of the same objects in arrays or other aggregations). But having that would make BeanDiffer to go from State.USEFUL_BUT_TRIVIAL to State.COOL. /O *********** REPLY SEPARATOR *********** On 2003-01-31 at 06:56 [EMAIL PROTECTED] wrote: > Well I definitely have some interest. In the past few days I have been > working on just this. The project that I'm working on is based on > 'Projects' which have a 'ProjectSpecification'. We have a need to maintain > an audit trail of changes to the ProjectSpecification, and the ability to > show the changes between 2 different ProjectSpecifications. The > ProjectSpecification consists of a fairly large (and growing) graph of > objects. > > I would certainly be interested in sharing ideas. > > Dave Derry > > > > > > > "Laird J. Nelson" > > <[EMAIL PROTECTED] To: > <[EMAIL PROTECTED]> > mherst.edu> cc: > > Subject: [BeanUtils] > Javabean diff? > 01/30/2003 07:23 PM > > Please respond to > > "Jakarta Commons > > Developers List" > > > > > > > > > Hi; I had written to the -user list about this subject a couple of days > ago, > and Henri Yandell expressed some interest, so I'll carry the nascent > thread > over here. > > I have a weird requirement at my real job related to Java bean graphs. > For > reasons that owe much more to bad web services design than anything else, > I > find I need to take two Java bean graphs--one which started out as a clone > of the other--and extract in some fashion only those paths that reflect > how > the second graph differs from the first. Think of the diff program > applied > to Java beans: the output, or a derivative of that output, would describe > only what modifications would need to be made to the first graph to make > it > become the second graph. > > I figured that this would make an interesting add-on to BeanUtils. There > also seem to be some JXPath-ish bits in here. Is there interest in the > BeanUtils/commons community for such a facility? > > Cheers, > Laird > -- > [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] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]