commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: [BeanUtils][Betwixt][commons] Proposal: Reflection/Introspection/MetaData package
Date Sun, 16 Jun 2002 15:06:39 GMT
As does Velocity....I think it calls it UberIntrospector or something.....
dIon Gillard, Multitask Consulting

Martin van den Bemt <>
06/17/02 12:19 AM
Please respond to "Jakarta Commons Developers List"

        To:     Jakarta Commons Developers List <>
        Subject:        Re: [BeanUtils][Betwixt][commons] Proposal: 
Reflection/Introspection/MetaData package

Also check this with the ant team, who have a lot of introspection in
there code.. It works on all jdk versions afaik. 

+1 on the idea btw..


On Sun, 2002-06-16 at 13:40, Stephen Colebourne wrote:
> Hi,
> Currently, Betwixt and other users of BeanUtils rely on the java.beans 
> Introspector to extract details from a class. Introspector is a very old 
> limited class in todays terms:
> - it doesn't support collections, just simple objects and arrays
> - it doesn't support modern conventions such as addXxx() adds to a the 
> list
> - it doesn't support overloading well
> - the bean info technique is difficult to code, poorly understood and
> limiting
> - it's just too plain dumb.
> I propose that BeanUtils/Betwixt/commons should replace Introspector 
with a
> more general purpose reflection/introspection package. Architecturally 
> would sit above reflection, but below Introspection:
> Requirements/Goals:
> - handle classes other than beans
> - support extensible metadata (not just for GUI builders)
> - handle normal (today) bean conventions (get/set/add/put methods)
> - handle future conventions that are not yet standard
> - support method overloading
> - be easily used directly from BeanUtils and Betwixt (and probably 
> - be a complete alternative to using java.lang.reflect
> - return immutable objects
> My proposed solution (not coded, fully open to discussion):
> Build a system with similarities to Digester. Rules get called when the
> class is examined determine how to link the methods together. For 
> the FindGetPropertyMethodRule would look at method names starting with 
> etc. The rule then classifies the method as a GET method and stores it 
> a structure something like this:
> - MethodSetInfo - holds details about a related set of methods.
> public String getName()
> public List getMethodInfos()
> public MethodInfo getMethodInfo(name)
> public Map getMetaData()
> public MetaData getMetaData(String name)
> - ClassInfo - main class that holds the representation of class. 
Subclass of
> MethodSetInfo
> public List getMethodSetInfos()
> public MethodSetInfo getMethodSetInfo(name)
> public List getMethodSetInfos(methodSetType)
> public MethodSetInfo getMethodSetInfo(methodSetType, name)
> public PropertyInfo getPropertyInfo(name)  // convenience
> - PropertyInfo - subclass of MethodSetInfo for properties (Lists/Maps 
> tbd)
> public Class getPropertyType()
> public MethodInfo getGetMethodInfo()  // convenience
> public MethodInfo getSetMethodInfo()  // convenience
> public Object getValue()
> public void setValue()
> - MethodInfo - categorised info about a method
> public String getName()
> public Method getMethod()
> publc String getCategory()
> public Map getMetaData()
> public MetaData getMetaData(String name)
> public Object invoke(object, args, respectAccessFlags)
> Attached to each element is the ability to hold MetaData. This is
> particularly important for Betwixt. It would allow the XMLBeanInfo class 
> be held directly on the representation of the class. And I'm sure other
> projects want MetaData - its supposed to be a long standing request for 
> (I know I need it for the Joda project).
> Note that I haven't expanded on the Rule part at the moment. Basically,
> people must be able to write their own rules and add them to the 
> rules for beans.
> As for which project it belongs with...I suggest lang or something new
> (reflect?). I would like to extract it from BeanUtils because its not 
> specific.
> Well, its an idea at the moment. There are some similarities to 
> but I think it goes way further. I already have a partial version of 
> but its specific to my needs. It needs some rework anyway, so I thought
> about if it could be generic. Opinions??
> Stephen
> PS. Three more possibilities for lang:
> Nameable
> MetaData
> Rule
> --
> To unsubscribe, e-mail:   <>
> For additional commands, e-mail: <>

To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

View raw message