Here is my plan:
All XML files and the annotated modules are parsed by the hivemind libs.
This can be executed by calling RegistryBuilder#autoDetectModules
or by manually loading the modules like this:
RegistryDefinition registryDefinition = new RegistryDefinition();
XmlModuleReader reader = new XmlModuleReader(registryDefinition);
reader.readClassPathModule("META-INF/hivemodule.xml");
reader.readClassPathModule("foo/hivemodule2.xml");
AnnotatedModuleReader reader = new
AnnotatedModuleReader(registryDefinition);
reader.readModules("**/*Module.class");
The resulting RegistryDefinition can be used to iterate over all
module definitions, its extension points and extensions.
(see
http://hivemind.apache.org/hivemind2/hivemind-framework/apidocs/org/apache/hivemind/definition/package-frame.html)
By iterating over the definitions a DOM tree gets created (like
RegistrySerializer does today)
and serialized. The resulting XML file does not conform to the hivemind
module deployment descriptor schema.
Finally a stylesheet transformation is applied on it to generate HTML.
Sounds straightforward but there are some catches:
* The core framework doesn't know anything about xml.
As a consequence, the ServicePointDefinition for example doesn't have a
schema
property (for service factories). But the information can be acquired
by casting a service point definition from a xml module to
XmlServicePointDefinitionImpl,
which gives access to the schema. So the plugin must be aware of the
different
definition types. In case of configuration points there is an additional
layer of indirection in form of ConfigurationParsers.
* configurations can be of any type and they needn't have an associated
schema.
Question is, how do you represent contributions which are not backed by
xml?
In the case of annotated modules it's not even possible to access the
contributed
data without constructing the registry.
* To read the annotated modules with AnnotatedModuleReader the module
classes
must be loaded. So the classpath must contain all referenced classes.
Whereas a maven plugin has access to the classpath which was used for
compilation
and is considered to be complete, hasn't it?
Achim
Am Tue, 10 Apr 2007 14:50:09 +0200 schrieb <[EMAIL PROTECTED]>:
Hi Johan,
I think, I used the Apache License - so no problem here.
However, I have to blushingly admit, that the hivedoc plugin is quite a
quick hack. So don't expect any clean abstraction from the
xml-Registry-Definition.
On the positive side, the code is short and straightforward, so it
shouldn't be really difficult to pull apart the xml-parsing and the
output-generation.
Unfortunately, I've not yet had the time to look into the new API.
Basically, one probably needs to search the source-files for
HiveMind-2.0 annotations.
This annotation-parsing thing should be common for validator and hivedoc
mojos. The simplest thing might be to generate the good old xml-registry
as a temporary file from the annotations and use it as input for both
tools.
Marcus
-----Original Message-----
From: Johan Lindquist [mailto:[EMAIL PROTECTED]
Sent: Tuesday, April 10, 2007 9:04 AM
To: [email protected]
Subject: Re: HiveMind Site
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi,
For the verifier mojo, there are no issues - not sure about
the HiveDoc one - Marcus?
And yes, HiveMind 2.0 integration should be considered.
verifier should definately be possibly to re-use for XML
descriptors but it needs to be retrofitted to handle
annotated registries. For HiveDoc, I would assume the same
thing applies?
Johan
Achim Hügen wrote:
> It would be great to add that plugin to the hivemind subversion
> repository as an additional module.
> Does any licensing issues or copyright speak against it?
>
> BTW, we will have to check how such plugins can be
integrated with the
> HiveMind 2 registry definition API.
>
> Achim
>
> Am Thu, 05 Apr 2007 10:09:17 +0200 schrieb Johan Lindquist
> <[EMAIL PROTECTED]>:
>
> Hi,
>
> Ok, I will take a look - and file the JIRA.
>
> About the plugin - basically there are two mojos that were
mentioned
> on the list - a hivedoc report plugin (written by Marcus
Schulte) and
> a hivemodule.xml verifier plugin (started by myself). The thoughts
> were to create one single hivemind plugin with these two
sub-goals (if
> that is the correct wording). Then of course, any other sub-goals
> that come along would slot in quite nicely as well. Makes sense?
>
> Cheers,
>
> Johan
>
> Achim Hügen wrote:
>>>> Your changes are very welcome.
>>>> Please use JIRA for submitting the patches.
>>>> Maybe the IE bug has already been fixed on the tapestry
site where
>>>> I copied the styles from.
>>>>
>>>> Could you give me a short update on the plugin subject?
>>>> Did you create the plugins or who is the owner?
>>>>
>>>> Achim
>>>>
>>>>
>>>> Am Tue, 03 Apr 2007 09:42:24 +0200 schrieb Johan Lindquist
>>>> <[EMAIL PROTECTED]>:
>>>>
>>>> Hi Achim,
>>>>
>>>> I think misunderstood the problem with the pom - thought it was
>>>> related to the "root" source repository location as indicated by
>>>> the project info on the homepage. Apologies for my
hasty conclusions.
>>>>
>>>> In any case, I think we all agree that the site should
be updated
>>>> and I understand fully if time prevents this . And as far as
>>>> committers go, I guess James is only one still active?
>>>>
>>>> On a different note - looks like the site doesn't work entirely
>>>> correct on IE (at least for me, the left nav is not
visible in IE -
>>>> firefox works fine) - do you mind if I make one or two
changes to
>>>> the hivemind site project to correct this? Would you
want another
>>>> JIRA for this to keep track?
>>>>
>>>> And lastly, was also hoping to incorpororate the Maven Hivemind
>>>> plugins previously mentioned - module verification and
hivedoc. Thoughts?
>>>>
>>>> Cheers,
>>>>
>>>> Johan
>>>>
>>>> Achim Hügen wrote:
>>>>>>> Johan, the bug refered to the HiveMind 2 poms and I fixed it.
>>>>>>> Nevertheless it would be nice if someone could apply
the patches
>>>>>>> to the 1.2 site and release it.
>>>>>>> Unfortunately my time is limited and I focus on
version 2.0 in
>>>>>>> the moment.
>>>>>>>
>>>>>>> Anybody else?
>>>>>>>
>>>>>>> Achim
>>>>>>>
>>>>>>> Am Mon, 02 Apr 2007 09:17:01 +0200 schrieb Johan Lindquist
>>>>>>> <[EMAIL PROTECTED]>:
>>>>>>>
>>>>>>> Hi All,
>>>>>>>
>>>>>>> Over the weekend there was a JIRA opened for the SVN links on
>>>>>>> the HiveMind site. Looks like the current project info (as
>>>>>>> generated by
>>>>>>> Maven) uses the site scm URL (someone correct me if I
am wrong)?
>>>>>>>
>>>>>>> Now, to make sure that users don't get confused can I
suggest we
>>>>>>> try to get the updated site up for Hivemind? The
patch already
>>>>>>> supplied will provide the new look and feel and other updates
>>>>>>> that the maven build brings. The 2.0 site should of course
>>>>>>> still be there but I guess it
>>>>>>> depends somewhat on how complete the documentation is
- Achim? If
>>>>>>> there is some work required to complete it, anything someone
>>>>>>> else can pick up?
>>>>>>>
>>>>>>> In any case, we should try to clean up the site and make sure
>>>>>>> links (such as the SCM) works correctly.
>>>>>>>
>>>>>>> Only my two cents worth, but we don't want to confuse
the users
>>>>>>> further by having the (so far very good) documentation fall
>>>>>>> apart on us.
>>>>>>>
>>>>>>> Thoughts?
>>>>>>>
>>>>>>> Cheers,
>>>>>>>
>>>>>>> Johan
>>>>>>>
>>>>
>
- --
you too?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFGGzbz1Tv8wj7aQ34RAp4iAJ9PpmaxZZdTZG3F/xFC1f9U92xO/ACeN2YR
mj/JzfMNeUYruPhBqbK4BYA=
=mhaO
-----END PGP SIGNATURE-----
