On 9/19/05, Kr <[EMAIL PROTECTED]> wrote: > Hi Robert, > > Your site is very interesting. I have packaged class and how do I feed > it to your tool at http://people.apache.org/~rdonkin/tool.html ? since > even if I keep my class in classpath your tool is not able to > recognize it .
the code is just intended to work as a design aid (we were talking about various approaches to developing mappings visually) rather than as functional software. the applets are part of an import phase where basic reusable mappings would be created for objects and then saved into a library. a second mapping phase would allow advanced mappings to be developed against given schema and is described in other design documents. the problem with the prototype is that it uses betwixt (which - as a dynamic binder - is well suited to fast prototyping) and has no output facility so it's just a toy. i am keen to try to create a proper application along the lines discussed but there's quite a bit of work that's needed. if it happens, it'll be open sourced somewhere (though not sure exactly where). be a while before it'd be ready for users. (be a bit different if you were interested in helping to develop it.) > Also how the tool take care of java collections like > java.util.Hashtable's i call these collective mappings. a key question is how (and whether) they are typed. betwixt is bean-centric. it extends the concept of properties to collective properties. adders are guessing using plural stemming (getItems/addItem). this allows a type to be guessed. jibx does less guessing (since it is field-centric) so you specify everything. see http://jibx.sourceforge.net/tutorial/binding-collects.html. more thought is needed for collections with arbitrary contents. not very easy to support (need a dynamic binder) and quite probably not really worth the effort. polymorphic collections (multiple specified sub-types) are more reasonable. > and java.util.HashMap' etc.? jibx takes the very reasonable approach that there is not natural mapping. betwixt guesses a reasonable mapping and tries to match a putter to a getter. this is the area where a good tool would be invaluable: it's a lot of work to have to specify in detail (ala jibx) but there isn't a single natural strategy. the problem with betwixt's introspection is that though the strategies make it flexible, they are too complex to learn easily. a visual tool could show the likely results and improve the speed at which mappings could be created (whether through jibx or betwixt). > Please execuse me > if my knowledge is trivial. it's not trivial: start from java has more than a few wrinkles and there are not very people who know much about it. - robert
