On 8/13/06, soumadeep sen <[EMAIL PROTECTED]> wrote:
Exactly, that's what I have been thinking too.
Cool :) Let me work on it a bit more. I was thinking whether to introduce a
scripting or rules engine support such as Groovy or Drools to handle the rules, your thoughts.
Note that spring has built-in support for Groovy. See http://static.springframework.org/spring/docs/2.0.x/reference/dynamic-language.html#dynamic-language-beans-groovy I was also thinking that the BAM component could leverage a rule engine features such as Drools, but i have not tought about it more yet. You can take a look at the Drools lightweight commponent, it may be helpfull. Yeah, I agree about the EIP component-wiretap. Let me also utilize it. Note that if you put a WireTap in front of the BAM component, the NMR will already process this message asynchronously without delaying the main exchange processing and the BAM component would not need to have an internal thread pool to process the exchanges asynchronously (it would only receive InOnly exchanges to process). Though i'm still not sure if this is the best way to go, as you would need a WireTap endpoint in front of all your BAM endpoints. The other way would be to include the eip component jar and inherit the WireTap to reuse the code as much as possible. Soumadeep
On 8/13/06, Guillaume Nodet <[EMAIL PROTECTED]> wrote: > > If the xml content is a stream, you first need to transform it to a > re-readable > xml source before any processing can be applied. > The EIP component does this and still give a NormalizedMessage > (take a look at it: oa.s.eip.support.MessageUtil). > You should really take a look at the wire tap implementation, as this is > the exact same thing that the BAM component does. > > If the Rule does not use a simple interface, the component > has to handle all your 16 cases internally. > I would prefer an extensible design, where the component would > first implement the xpath rule on the content and where rules > could be written by users for custom things. And as you say, > even a test on a property could have lots of different implementations: > someone may want to test a string property which will match > a regular expression, or an integer > value, ... > > I think all these cases can not be handled by the component, without > leading to a bad xml schema and a code which will be difficult to > maintain. > > > On 8/12/06, Soumadeep-Infravio <[EMAIL PROTECTED]> wrote: > > > > I actually agree with you in all accounts but just trying to decouple > the > > config sources so that there is minimum dependency if it's fed from > > external > > sources. But either way the internal dependency will be there. > > > > One important thing that I need to mention here is that you had > commented > > about having an interface for what to evaluate, such as > properties,content > > etc... Let's discuss about this as the scope will broaden a lot: > > > > The possible sources that one could monitor are: > > From the MessageExchange: (These in my opinion are not relevant as they > > are > > more application specific and not content specific) > > 1) Endpoint 2) Error 3) ExchangeId 4) Fault 5) InterfaceName 6) > Operation > > 7) > > Pattern > > 8) Property 9) Role 10) Service 11) Status 12) IsTransacted 13) Message > > > > For (13) Message alone could have > > 13.1) Attachment (have to deal with DataHandler) > > 13.2) Property (handle Object) > > 13.3) SecuritySubject ( subject and principal) > > 13.4) Content (xml- being handled now) > > > > The rule definition will change as each of the above needs to be > uniquely > > handled. This will get pretty complicated as we may then have to have an > > interface for the Rule and have different implementations, example -> > > XPathRule could be an implementation where we can handle it based on > what > > we > > have, PropertyBasedRule would introduce another story where the > evaluation > > could be expression based, With attachment it could be anything - > depends > > on > > the DahaHandler... > > > > So rather than having Object evaluate(NormalizedMessage message) we > could > > have an interface for the BAMProcessor and pass appropriate values to > the > > implementation. The other thing is we can't pass the NormalizedMessage > > directly to the processor as it's an async operation so we should not > wait > > for any values to be passed from the processor. It will make it > blocking. > > (as in the processor we may take the content out so there would be no > > content left after delegation to the processor) > > > > > > Your thoughts... > > > > Soumadeep > > > > ----- Original Message ----- > > From: "Guillaume Nodet" <[EMAIL PROTECTED]> > > To: <servicemix-dev@geronimo.apache.org> > > Sent: Saturday, August 12, 2006 2:36 PM > > Subject: Re: [jira] Created: (SM-534) Business Activity Monitoring > > Component > > > > > > > On 8/11/06, Soumadeep-Infravio <[EMAIL PROTECTED]> wrote: > > >> > > >> Thanks for the reply, my comments are inline... > > >> > > >> ----- Original Message ----- > > >> From: "Guillaume Nodet" <[EMAIL PROTECTED]> > > >> To: <servicemix-dev@geronimo.apache.org> > > >> Sent: Friday, August 11, 2006 3:02 PM > > >> Subject: Re: [jira] Created: (SM-534) Business Activity Monitoring > > >> Component > > >> > > >> > > >> >A few comments: > > >> > > > >> > 1) Files should include the Apache standard header > > >> > > >> Will include them > > >> > > >> > 2) resources are loaded with xbean in BAMEndpoint.process > > >> > they override any definition specified directly with the > rules, > > >> > actions properties > > >> > IMHO, they should be loaded when activate is called (or at > > >> > initialization time, > > >> > by implementing the spring interface InitializingBean) and > only > > if > > >> the > > >> > properties > > >> > are not set > > >> > > >> I had kept both ways to load properties, but going forward I think > > >> resource based file loading would be the most appropriate solution so > > >> will > > >> keep that and remove the embedded beans defn. from BAMEndpoint. Let > me > > >> know > > >> if I missing the point. > > > > > > > > > Mmmh, i guess it depend. Having to create 4 xml files instead of 1 is > > not > > > always the easiest way ... > > > > > >> 3) I don't see any use of the BAMGlobalConfig / Params classes > > >> > > >> This has been kept for future use where global parameters can be > > defined > > >> and used by adaptors as well as actions and rules > > >> > > >> > 4) Rules are not extensible. People will need to use to be able to > > >> check > > >> > for > > >> > properties, attachments, not only xpath on the content. So it > > >> > should > > >> be > > >> > an interface with > > >> > Object evaluate(MessageExchange exchange) > > >> > or > > >> > Object evaluate(NormalizedMessage message) > > >> > or anything like that. > > >> > > >> This is a good idea, for now let me just implement for content based > > then > > >> we can provide other adaptors later > > >> > > >> > 5) When using xpath, you need a way to configure the namesapces in > > use > > >> in > > >> > the xpath > > >> > expression, else you can not use it on xml requests with > > namespaces. > > >> > Take a look at the XPathPredicate in servicemix-eip > > >> > > >> Will make sure that it handles namespaces > > >> > > >> > 6) The configuration could leverage much more of spring/xbean > > features > > >> and > > >> > use a clean POJO model > > >> > which will be easily translated into a clean xml schema using > > >> > xbean. > > >> > They don't need to use id references and class names, as spring > > >> > will > > >> do > > >> > that easily. > > >> > And instead of using BAMActionParameters, these parameters > could > > be > > >> > easily configured on > > >> > the action / adaptor itself. > > >> > > > >> > So instead of having > > >> > <bam:bAMRule description="Email rule1" resultType="Boolean" > > >> > ruleName="emailRule1" xpath="/[EMAIL PROTECTED]'555-3482']"> > > >> > <bam:actionDetails> > > >> > <bam:bAMActionDetail executeOn="true" actionID="printer" > > >> > active="true"/> > > >> > </bam:actionDetails> > > >> > </bam:bAMRule> > > >> > <bam:bAMAction actionName="printer" adaptorClass=" > > >> > org.apache.servicemix.bam.sample.PrintAdaptor" description="For > > >> emailing"> > > >> > <bam:params> > > >> > <bam:bAMActionParameter name="mobileNumber" > > value="405-3785" > > >> > type="String"/> > > >> > <bam:bAMActionParameter name="address" value="680, morse > > ave > > >> CA" > > >> > type="String"/> > > >> > </bam:params> > > >> > </bam:bAMAction> > > >> > > > >> > You could simply have > > >> > <bam:endpoint ...> > > >> > <bam:rule> > > >> > <bam:evaluator> > > >> > <bam:xpath xpath="/test:sample/@id > > " nsContext="#nsContext" > > >> /> > > >> > </bam:evaluator> > > >> > <bam:actions> > > >> > <bam:action executeOn="554-345" adaptor="#printAdaptor" > > /> > > >> > </bam:actions> > > >> > </bam:endpoint> > > >> > > > >> > <bam:print id="printAdaptor" output="stderr" > > mobileNumber="405-3785" > > >> > address="680, morse ave CA" /> > > >> > > > >> > <bam:namespaceContext id="nsContext"> > > >> > <bam:namespace prefix="test">http://test</bam:namespace> > > >> > </bam:namespaceContext> > > >> > > > >> > This is only an example to show how to use references, without > having > > >> > to > > >> > define IDs and classNames. > > >> > This lead to a much cleaner POJO model. > > >> > > >> The reason it has been repesented this way is because an action can > be > > >> referenced by multiple rules. So you can define an action once and > then > > >> reference it from different rules.This provides separation of action > > from > > >> rules. > > >> > > >> The relationship is as follows:(haven't included the attributes) > > >> > > >> BAMAction has one or more BAMActionParameter (bamActionParameter has > > >> details of the parameters that the action adaptor would use) > > >> BAMRule has one or more BAMActionDetail (bamActionDetail has the > > >> reference > > >> to actions using an ID) > > >> > > >> So from a user perspective in the rule.xml file they can specify an > > >> actionid which is there in the actions.xml . so even if the action > > >> implementation changes the rules file doesn't need to be changed. > > >> > > >> Consider a situation where these files could be coming from an > external > > >> source so rather than getting a whole bulk of xml we can isolate and > > deal > > >> with specific xmls. > > >> > > >> From an dev perspective, it's more of a parent child relationship-> > > later > > >> on storage could be easy in a database maybe. > > >> > > >> I have used BAMActionParameter as well as BAMGlobalParameter -> they > > both > > >> are same so I guess, use one class and name it BAMParameter. > > > > > > > > > I do understand the relationship between the classes. What i meant > is > > > that > > > instead of implemeting the > > > wiring inside the BAM component, you could leverage > > spring features: you > > > can reference beans using their > > > ids and share them. The benefit is that you can have beans with > > > constructors and you can inject properties > > > on them, instead of only relying on a map of strings to do that. It' > s > > > just > > > more powerfull, but you could still > > > use IDs. > > > > > > Note that in spring, you can also import other spring files, so you > > could > > > just have one main spring file (the one containing the endpoint) , > > import > > > the other ones and use references on these imported beans. > > > > > > It' s just about leveraging spring features ;) > > > > > > My $0.02 :-) > > >> > > >> > > >> > > > >> > > > >> > > > >> > On 8/10/06, Soumadeep Sen (JIRA) <[EMAIL PROTECTED]> wrote: > > >> >> > > >> >> Business Activity Monitoring Component > > >> >> -------------------------------------- > > >> >> > > >> >> Key: SM-534 > > >> >> URL: > > https://issues.apache.org/activemq/browse/SM-534 > > >> >> Project: ServiceMix > > >> >> Issue Type: New Feature > > >> >> Components: servicemix-common > > >> >> Reporter: Soumadeep Sen > > >> >> Attachments: servicemix-bam.zip > > >> >> > > >> >> This Business Activity Monitoring component which works off an > xpath > > >> >> expression. The xpath expression acts as a Key performance > > indicator. > > >> Based > > >> >> on the xpath evaluation, actions can be triggered. These actions > can > > >> >> be > > >> >> implemented by users by extending the BAMAdaptor interface's > execute > > >> method > > >> >> which takes an array of BAMActionParameter. > > >> >> > > >> >> For providing details in terms of Rules,Actions and global > > parameters, > > >> the > > >> >> actions.xml, rules.xml and globalConfig.xml need to be populated. > > >> Sample > > >> >> files can be found in the src/test/resources directory and usage > > >> details for > > >> >> the BAMComponent can be found in the spring.xml file which is in > the > > >> same > > >> >> directory. > > >> >> > > >> >> The relationship between actions and rules is established by id > > >> reference > > >> >> where in the rules have actions IDs (no or more). Please refer the > > >> >> respective xml files. > > >> >> > > >> >> All implemented adaptor classes could be put in the option lib > > >> directory > > >> >> of smx so that they can be referenced by the BAM processor. (A > > sample > > >> >> Adaptor called PrintAdaptor has been provided in the src...samples > > dir > > >> for > > >> >> reference) > > >> >> > > >> >> Will be putting a wiki page shortly which will have more details. > > >> >> > > >> >> > > >> >> > > >> >> > > >> >> > > >> >> > > >> >> -- > > >> >> This message is automatically generated by JIRA. > > >> >> - > > >> >> If you think it was sent incorrectly contact one of the > > >> >> administrators: > > >> >> https://issues.apache.org/activemq/secure/Administrators.jspa > > >> >> - > > >> >> For more information on JIRA, see: > > >> http://www.atlassian.com/software/jira > > >> >> > > >> >> > > >> >> > > >> > > > >> > > > >> > -- > > >> > Cheers, > > >> > Guillaume Nodet > > >> > > > >> > > > > > > > > > > > > -- > > > Cheers, > > > Guillaume Nodet > > > > > > > > > > > > -- > Cheers, > Guillaume Nodet > >
-- Cheers, Guillaume Nodet
