Hi Simon,
I dont think using an underlying tuscany jdk logger would be useful to
plugins as they may not want to log, rather show it somewhere else such as
console etc. Tuscany can use an underlying logger in it's own monitor ( as
it uses today). But i think the first approach of using a monitor is better
along with the condition that it be made more usable by the plugins by
giving them greater control.
Another point is that tuscany should use ResourceBundle for validation
messages as well. I dont think this is being done today.
regards
Hasan
On Wed, Apr 9, 2008 at 1:22 PM, Simon Laws <[EMAIL PROTECTED]>
wrote:
> On Wed, Apr 9, 2008 at 12:49 PM, Simon Laws <[EMAIL PROTECTED]>
> wrote:
>
> >
> >
> > On Wed, Apr 9, 2008 at 12:00 PM, Hasan Muhammad <[EMAIL PROTECTED]>
> wrote:
> >
> > > Hi Simon,
> > >
> > > I am on revision 634808. The ContributionServiceImpl has changed since
> > > then,
> > > and with the one that i have, it would lead through the
> > > CompositeProcessor
> > > instead of the CompositeDocumentProcessor. Hence the difference in
> > > exceptions..
> > >
> > > Also, dont you think that with the error that you got should throw an
> > > exception with schema validation, rather than just a warning?
> > >
> > > Hasan
> > >
> > > On Wed, Apr 9, 2008 at 6:36 AM, Simon Laws <[EMAIL PROTECTED]>
> > > wrote:
> > >
> > > > On Tue, Apr 8, 2008 at 2:58 PM, Hasan Muhammad <[EMAIL PROTECTED]>
> > > wrote:
> > > >
> > > > > Hi Simon,
> > > > >
> > > > > Thank you for the good information. First up i am trying to verify
> > > > whether
> > > > > the schema validation works when we point to our schemas. Can you
> > > let me
> > > > > know what is a simple error that i can introduce so that i can
> > > verify
> > > > > this?
> > > > > I tried doing this to my composite file (In block red):
> > > > >
> > > > > <component name="MyServiceComponentNew">
> > > > > <implementation.java
> > > > class="mysca.test.myservice.impl.MyServiceImpl"/>
> > > > > *<binding.ws/>*
> > > > > <property name="location" source="$newLocation"/>
> > > > > <property name="year" source="$newYear"/>
> > > > > </component>
> > > > >
> > > > > This resulted in the following exception, but i think this is part
> > > of
> > > > the
> > > > > validation done by artifact processor and would result even if we
> > > > comment
> > > > > out the schema validation.
> > > > >
> > > > >
> > > org.apache.tuscany.sca.contribution.service.ContributionReadException:
> > > > > Unexpected <binding> element found. It should appear inside a
> > > <service>
> > > > or
> > > > > <reference> element
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.assembly.xml.CompositeProcessor.read(CompositeProcessor.java:373)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.assembly.xml.CompositeProcessor.read(CompositeProcessor.java:75)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.contribution.processor.ExtensibleStAXArtifactProcessor.read(ExtensibleStAXArtifactProcessor.java:83)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.contribution.service.impl.ContributionServiceImpl.processReadPhase(ContributionServiceImpl.java:475)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.contribution.service.impl.ContributionServiceImpl.addContribution(ContributionServiceImpl.java:383)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> org.apache.tuscany.sca.contribution.service.impl.ContributionServiceImpl.contribute(ContributionServiceImpl.java:202)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.soa.sca.runtime.impl.DomainCompositeHelper.addContribution(DomainCompositeHelper.java:75)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.soa.sca.runtime.impl.SCAContainerComponentImpl.startComposite(SCAContainerComponentImpl.java:235)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.soa.sca.admin.runtime.tuscany.SCATuscanyRuntimeHandlerImpl.startModule(SCATuscanyRuntimeHandlerImpl.java:125)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.soa.sca.admin.runtime.impl.SCARuntimeImpl.start(SCARuntimeImpl.java:349)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.soa.sca.admin.runtime.impl.SCARuntimeImpl.start(SCARuntimeImpl.java:446)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.runtime.component.CompositionUnitMgrImpl.start(CompositionUnitMgrImpl.java:331)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.runtime.component.CompositionUnitImpl.start(CompositionUnitImpl.java:126)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.runtime.component.CompositionUnitMgrImpl.start(CompositionUnitMgrImpl.java:281)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.ws.runtime.component.CompositionUnitMgrImpl$CUInitializer.run(CompositionUnitMgrImpl.java:768)
> > > > > at
> > > > >
> > > > >
> > > >
> > >
> com.ibm.wsspi.runtime.component.WsComponentImpl$_AsynchInitializer.run(WsComponentImpl.java:348)
> > > > > at com.ibm.ws.util.ThreadPool$Worker.run(ThreadPool.java:1487)
> > > > >
> > > > >
> > > > > regards
> > > > >
> > > > > On Tue, Apr 8, 2008 at 5:56 AM, Simon Laws <
> > > [EMAIL PROTECTED]>
> > > > > wrote:
> > > > >
> > > > > > On Mon, Apr 7, 2008 at 4:33 PM, Hasan Muhammad <[EMAIL PROTECTED]
> >
> > > > wrote:
> > > > > >
> > > > > > > Currently, i see that we have various places where we can plug
> > > in
> > > > > > > validation
> > > > > > > monitoring. Some of the ones that i found are in the
> > > > > > > ReallySmallRuntimeBuilder as shown below:
> > > > > > >
> > > > > > > public static CompositeBuilder
> > > > > createCompositeBuilder(AssemblyFactory
> > > > > > > assemblyFactory,
> > > > > > >
> > > > > > SCABindingFactory
> > > > > > > scaBindingFactory,
> > > > > > >
> > > > > > > IntentAttachPointTypeFactory intentAttachPointTypeFactory,
> > > > > > >
> > > > > > > InterfaceContractMapper interfaceContractMapper,
> > > > > > >
> > > > > List<PolicySet>
> > > > > > > domainPolicySets) {
> > > > > > > return new CompositeBuilderImpl(assemblyFactory,
> > > > > > scaBindingFactory,
> > > > > > > intentAttachPointTypeFactory, interfaceContractMapper,
> > > > > domainPolicySets,
> > > > > > > null);
> > > > > > > }
> > > > > > >
> > > > > > > public static DomainBuilder
> > > createDomainBuilder(AssemblyFactory
> > > > > > > assemblyFactory,
> > > > > > > SCABindingFactory scaBindingFactory,
> > > > > > > IntentAttachPointTypeFactory
> > > > intentAttachPointTypeFactory,
> > > > > > > InterfaceContractMapper interfaceContractMapper,
> > > > > > > List<PolicySet> domainPolicySets) {
> > > > > > > return new DomainWireBuilderImpl(assemblyFactory,
> > > > > > > scaBindingFactory,
> > > > > > > intentAttachPointTypeFactory, interfaceContractMapper,
> > > > > domainPolicySets,
> > > > > > > null);
> > > > > > > }
> > > > > > >
> > > > > > > Instead of passing null, we can pass in our own
> > > > CompositeBuildMonitor.
> > > > > > Are
> > > > > > > these the only places where we need to do this? or are there
> > > other
> > > > > > places
> > > > > > > as
> > > > > > > well?
> > > > > > >
> > > > > > > Hasan
> > > > > > >
> > > > > >
> > > > > > Hi Hasan
> > > > > >
> > > > > > There are a few places that logically we need to be able to
> > > validate
> > > > > input
> > > > > > and hence add monitors. The story is changing a little as we are
> > > > moving
> > > > > > toward the new "workspace" code for managing contributions but
> > > > > logically
> > > > > > I
> > > > > > would expect to be running validating on at least the following
> > > > > >
> > > > > > - Contribution processing
> > > > > > - dependency (imports/exports) analysis
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> [org.apache.tuscany.sca.workspace.builder.impl.ContributionDependencyBuilderImpl]
> > > > > > - Composite read
> > > > > > - schema compliance
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> [org.apache.tuscany.sca.contribution.processor.DefaultValidatingXMLInputFactory
> > > > > > - only appears to be initialized in ReallySmallRuntimeBuilder so
> > > need
> > > > to
> > > > > > look at this in context of workspace]
> > > > > > - policy intent matching
> > > > > > [org.apache.tuscany.sca.assembly.xml.CompositeDocumentProcessor]
> > > > > > - extension availability [General artifact processor
> hierarchy]
> > > > > > - Composite resolve
> > > > > > - ensure that artifacts required by the composite are
> available
> > > > > [General
> > > > > > artifact resolver hierarchy]
> > > > > > - Composite build
> > > > > > - ensure that the composite is valid and consistent, e.g.
> unique
> > > > > > component names, valid reference targets etc.
> > > > > >
> > > [org.apache.tuscany.sca.assembly.builder.impl.CompositeBuilderImpl]
> > > > > >
> > > > > > I've noted in the square brackets where this function currently
> > > is.
> > > > Note
> > > > > > also that I'm making no comment here about whether the
> > > construction of
> > > > > the
> > > > > > in-memory composite model is for use purely for contribution
> > > > processing
> > > > > or
> > > > > > is going to activated and started in a runtime. This validation
> is
> > > > > > appropriate in both cases although you may choose to use
> different
> > > > > > monitors
> > > > > > in the two cases.
> > > > > >
> > > > > > So what I suggest as a first step is that you go ahead and
> change
> > > > > > ReallySmallRuntimeBuilder to plug a monitor into
> > > CompositeBuilderImpl
> > > > to
> > > > > > see
> > > > > > how it works. We can work here to build a consistent view of
> > > > > >
> > > > > > - all the places a monitor is required
> > > > > > - what should the plugin model for monitors be
> > > > > > - the flow of control (monitors vs exceptions)
> > > > > > - what extra features may be required, I18N etc.
> > > > > >
> > > > > > I'm going to try a few experiments too to familiarize myself
> with
> > > this
> > > > a
> > > > > > bit
> > > > > > more.
> > > > > >
> > > > > > Regards
> > > > > >
> > > > > > Simon
> > > > > >
> > > > >
> > > >
> > > > Hi Hasan
> > > >
> > > > The change you suggest should cause a schema validation error. I
> made
> > > the
> > > > change locally in samples/calculator and see the following output in
> > > > stdout
> > > >
> > > > 08-Apr-2008 22:41:54
> > > >
> > >
> org.apache.tuscany.sca.contribution.processor.ValidatingXMLStreamReader$1
> > > > error
> > > > WARNING: XMLSchema validation problem in: null, line: 28, column: 5
> > > > cvc-complex-type.2.4.a: Invalid content was found starting with
> > > element '
> > > > binding.ws'. One of '{"http://www.osoa.org/xmlns/sca/1.0":include, "
> > > > http://www.osoa.org/xmlns/sca/1.0":service, "
> > > > http://www.osoa.org/xmlns/sca/1.0":property, "
> > > > http://www.osoa.org/xmlns/sca/1.0":component, "
> > > > http://www.osoa.org/xmlns/sca/1.0":reference, "
> > > > http://www.osoa.org/xmlns/sca/1.0":wire, WC[##other:"
> > > > http://www.osoa.org/xmlns/sca/1.0"]}' is expected.
> > > > 08-Apr-2008 22:41:54
> > > >
> > > >
> > >
> org.apache.tuscany.sca.contribution.processor.ExtensibleStAXArtifactProcessor
> > > > read
> > > > WARNING: Element
> > > > {http://www.osoa.org/xmlns/sca/1.0}binding.ws<http://www.osoa.org/xmlns/sca/1.0%7Dbinding.ws>
> <http://www.osoa.org/xmlns/sca/1.0%7Dbinding.ws>
> > > <http://www.osoa.org/xmlns/sca/1.0%7Dbinding.ws>cannot be
> > > > processed. ([row,col {unknown-source}]: [28,5])
> > > >
> > > > No exception is raised though.
> > > >
> > > > Let me take a look at what has changed.
> > > >
> > > > Simon
> > > >
> > >
> >
> > Ok, thanks for that Hasan. I see the change you are referencing. I'm not
> > sure why the exception is no longer thrown. The code is still in the
> > CompositeProcessor to trap the case but maybe the offending element is
> > removed from the stream because of the validation warning.
> >
> > I would argue that we shouldn't mandate that an exception is thrown in
> > this case but that the validation mechanism is opened up to the monitor
> API
> > so that a monitor can be attached and those who are embedding tuscany
> can
> > decide how they want to deal with errors like this.
> >
> > Simon
> >
> >
> >
> Hi
>
> I've just been through the original list again looking at what happens
> with
> validation types errors (by which I mean a workspace type admin console
> would reasonably want to display based on user input/changes). This is
> what
> I found.
>
> - Contribution processing
> - dependency (imports/exports) analysis
>
>
> org.apache.tuscany.sca.workspace.builder.impl.ContributionDependencyBuilderImpl
> uses a ContributionDependencyBuilderMonitor
> - Composite read
> - schema compliance
>
>
> org.apache.tuscany.sca.contribution.processor.DefaultValidatingXMLInputFactory
> Uses writes to local JDK Logger
> - policy intent matching
> org.apache.tuscany.sca.assembly.xml.CompositeDocumentProcessor
> Throws exceptions
> - extension availability
> General artifact processor hierarchy
> Throws exceptions
> - SCA compliance, e.g. annotation correctness
> General artifact processor hierarchy
> Throws exceptions
> - Composite resolve
> - ensure that artifacts required by the composite are available
> General artifact resolver hierarchy
> Throws exceptions
> - Composite build
> - ensure that the composite is valid and consistent, e.g. unique
> component names, valid reference targets etc.
> org.apache.tuscany.sca.assembly.builder.impl.CompositeBuilderImpl
> uses a CompositeBuilderMonitor
>
> I would like to look for a more consistent approach across these various
> elements of validation. The aim being to make it more obvious how to
> recognize validation type errors and act accordingly. There has been a
> relatively lengthy debate about how to group and report errors in [1]
> which
> is worth a read. The scenarios I have in mind are workspace administration
> type scenarios where, for example, A user may
>
> Add a contribution and be notified that contributions that it depends on
> are
> not available
> Look at a composite and be notified that there are XML and/or SCA errors
> etc.
>
> From [1] we want to capture as many errors as possible before reporting
> back
> to the user. Two things come immediately to mind.
>
> 1 - Pull up the Monitor API and make it more generic (and from previous
> comments here make it an extension point)
> 2 - Rely on a Tuscany specific JDK logger and remove the monitor (or at
> least don't extend it further).
>
> I'm going to think some more about these options. Any thoughts or more
> options?
>
> Simon
>
> [1] http://www.mail-archive.com/tuscany-dev@ws.apache.org/msg29878.html
>
