I've just committed a start of this, might be better to be in a scratch area to start with but it was easier this way, can maybe move it out later. The annotations are in org.apache.synapse.mediators.annotations and theres and AnnotatedCommandMediator, and a testcase in org.apache.synapse.mediators.ext, which has class AnnotatedCommand2 which shows this being used. So far only supports reading string values from the MessageContext but thats enough to help refine things.
I couldn't see a way to support any number of arbitrary namespace prefixes easily, so right now you can use @Namespace with a single arbitrary namespace prefix or with "ns", "ns1", "ns2"..."ns5". Thats on both the class and field/method so that enables defining two arbitrary namespace prefixes and if you need more than that you have to use the ns or ns1 to ns5 prefixes. Maybe someone can think of a better way of doing this? You have to define the soap enveloper and header or body tags in the xpath witch doesn't seem great. Maybe it would be better to have something like separate annotations for the headers and body. I'm also wondering again if it may be simpler to have a single annotation instead of the ReadFromMessage/ReadAndUpdate/UpdateMessage ones, and have the method being a getter/setter determine if its read or update and fields always do read and update. Anyway open to review, what to people think? ...ant On Nov 20, 2007 12:38 PM, Paul Fremantle <[EMAIL PROTECTED]> wrote: > Aha > > That is quite cunning. So in other words, if you annotate the field then > we assume to get it before execute and set it after. > > If you annotate the getter then we only get, annotate the setter we only > set. > > I like the model, but I'd like to make the annotations match the action. > > So: > @namespace(ns="http://fremantle.org); > @ReadFromMessage(xpath="/ns:quote/Symbol") > public void setSymbol(String symbol) { > > ... > > } > > @UpdateMessage(xpath="/") > public OMElement getPayload() { > // return an OMElement > } > > @ReadAndUpdate(xpath="blah") > String symbol = null; > > // expecting getters and setters: > String getSymbol() { } > void setSymbol(String s) { } > > Does that make sense? > > Since XPaths can logically refer to Strings, Booleans and Integers as well > as XML, I suggest we support those as property types in the class too. > > Paul > > > > On Nov 20, 2007 12:30 PM, ant elder <[EMAIL PROTECTED]> wrote: > > > I'm not sure it needs a getter/setter generated or the type attribute > > specified on the property annottaion. The @property annotation could be > > used on either a field or getter/setter method: > > > > @property(name="symbol") > > String value > > > > or > > > > @property(name="symbol") > > public void setValue(String s) { > > value = s; > > } > > > > or > > > > @property(name="symbol") > > public String getValue() { > > return value; > > } > > > > The annotation is associated with the field or method so the type can > > easily be introspected from that. > > > > Also, when the annotation is associated with a method you can see if its > > a getter or a setter so the action can be determined from the method name > > (get=out, set=in). > > > > ...ant > > > > > > On Nov 20, 2007 11:52 AM, Paul Fremantle <[EMAIL PROTECTED] > wrote: > > > > > Sorry that wasn't very clear was it! > > > > > > Basically, I thought one approach would be to add a name and type > > > parameter to the > > > @property tag > > > > > > @property(name="symbol", type="String|OMElement",....) > > > > > > and then (I'm assuming - based on my limited knowledge of annotations) > > > we could automatically generate getters and setters. > > > > > > The problem with this approach is that the getters/setters would not > > > be available for command completion in the IDE, so I ditched this idea. > > > > > > Paul > > > > > > > > > On Nov 20, 2007 11:47 AM, ant elder <[EMAIL PROTECTED] > wrote: > > > > > > > > > > > > > > > On Nov 20, 2007 11:44 AM, Paul Fremantle <[EMAIL PROTECTED]> wrote: > > > > > > > > <snip> > > > > > > > > - the action could really be optional as its not so hard for the > > > > > > runtime to see that the value has been changed and set/getandset > > > > > > would just > > > > > > be a performance optimisation > > > > > > > > > > I guess so. It depends on whether we generate the property and > > > > > getters/setters or not. I was kind of assuming that we wouldn't > > > > > generate > > > > > them. Alternatively we could cache values before and after the execute > > > > > method, but thats a bit yucky, I think its so simple to use an > > > > > annotation, > > > > > and also since you get command completion for annotations inside IDEs > > > > > we can > > > > > make it a required property. > > > > > > > > > > > > > > > > > > What do you mean by "generate the property and getters/setters"? > > > > > > > > ...ant > > > > > > > > > > > > > > > > -- > > > Paul Fremantle > > > Co-Founder and VP of Technical Sales, WSO2 > > > OASIS WS-RX TC Co-chair > > > > > > blog: http://pzf.fremantle.org > > > [EMAIL PROTECTED] > > > > > > "Oxygenating the Web Service Platform", www.wso2.com > > > > > > > > > > -- > Paul Fremantle > Co-Founder and VP of Technical Sales, WSO2 > OASIS WS-RX TC Co-chair > > blog: http://pzf.fremantle.org > [EMAIL PROTECTED] > > "Oxygenating the Web Service Platform", www.wso2.com >
