I had a lot more thoughts on this subject while working up my
apachecon eu talk on geronimo server assemblies.
Right now we have a _lot_ of deployer code that carefully examines
exactly what your app needs and depending on what it finds adds
dependencies to the classloader. For instance if it figures out you
have a web app and it's the jetty deployer it adds the jetty car as a
parent. (Some of the smaller deployer bits I think currently always
add their dependencies since there was a problem with changing the
classpath but I hope this can be fixed with the recent configuration/
classloader work).
I think the plugin archetype should do this too. You tell it what app
you are trying to deploy and some kind of profile like jetty/tomcat
and cxf/axis2 and it (should) use the same logic as the deployer to
figure out what is needed. For the archetype, it should be adding the
dependencies and deployer dependencies to the pom and listing the
deployer dependencies the car-maven-plugin needs to start.
re stringtemplate vs. velocity.... I don't know enough about how the
archetype framework to judge whether anything except velocity is
plausible.
Are you actively working on this project? If not I might see if
there's some fairly easy way to access the decision making the current
deployers use.
thanks!
david jencks
On Mar 17, 2009, at 1:26 AM, Shawn Jiang wrote:
It looks to me as if there are no plans in the javaee apps parts of
the samples apps, so the plans are all in the plugin modules. I
really don't think we want to duplicate the functionality of the
existing archetypes that do a fine job of creating projects for java
ee apps. If you really do want to help people make projects like
the sample apps, the way to do that is to improve the plugin
archetype.
Thanks for your comments. actually, what I said was to make the
pom.xml and plan.xml template but not the JEE app itself. For
example:
servlet-examples/
pom.xml
servlet-examples-jetty/
pom.xml
src/
main/
plan/
plan.xml
servlet-examples-tomcat/
servlet-examples-war/
From this pespective, what I was talking about is to create plugins
archetype instead of JEE app archetype.
I like your ideas about what can be generated in the geronimo
plan. I think we can also generate a lot more of the pom than we do
now.
I agree.
I would be very tempted to try stringtemplate instead of velocity.
I think its clear MVC separation and roots in code generation
technology (it's the output stage for antlr) may prove useful for
the manipulation we need to do.
But how could we reuse the existing powerful maven archetype if
using stringtemplate ? I feel that StringTemplate is more useful in
a MVC framework but not in a project generating tool. Do you agree ?
While enabling people to reproduce our sample apps is probably
better than what we do now, I'd prefer to help them with apps that
actually do a realistic amount of work. I think improving the
plugin archetype would be a really big help.
I suggest to start with the current sample plugins. The final goal
is to reduce most of the effort to release all samples for each
geronimo version by providing stable archetype.
thanks
david jencks
On Tue, Mar 17, 2009 at 3:35 PM, David Jencks
<[email protected]> wrote:
On Mar 16, 2009, at 11:50 PM, Shawn Jiang wrote:
From my point of view, every current geronimo sample could be
converted to a archetype. So that the user just needs install maven
and he could just type a command like this:
mvn archetype:generate
-DarcheTypeCatalog=http://geronimo.apache.org/archetypes/samples
he will get a sample menu like this:
Choose archetype:
1: internal -> hello (geronimo sample web application project)
2: internal -> calculator-stateless-pojo (geronimo sample ejb
project)
3: internal -> mdb (geronimo Message-Driven Bean sample project)
4: internal -> GBean(simple geronimo GBean project)
5: internal -> xxxx
6: internal -> xxxx
7: internal -> xxxx
Choose a number: (1/2/3/4/5/6/7) 1: : 1
select "1" to create a hello sample project will bring him to a
wizard like this:
Define value for geronimo_version: : 2.1.4
Define value for context_root: : hellocontext
Define value for groupId: : my.first.geronimo.web
Define value for artifactId: : helloApp
Define value for version: 1.0-SNAPSHOT: :
Define value for package: my.first.geronimo.web: :
Confirm properties configuration:
geronimo_version: 2.1.4
context_root: hellocontext
groupId: my.first.geronimo.web
artifactId: helloApp
version: 1.0-SNAPSHOT
package: my.first.geronimo.web
Y: : Y
notice the user could define which geronimo version he is using and
defind the context root when generating the application. The two
values will be used to generate a geronimo-web.xml like this:
<web-app xmlns="http://geronimo.apache.org/xml/ns/j2ee/web-2.0.1";
xmlns:naming="http://geronimo.apache.org/xml/ns/naming-1.2";
xmlns:sec="http://geronimo.apache.org/xml/ns/security-2.0";
xmlns:sys="http://geronimo.apache.org/xml/ns/
deployment-1.2">
<sys:environment>
<sys:moduleId>
<sys:groupId>my.first.geronimo.web</sys:groupId>
<sys:artifactId>helloApp</sys:artifactId>
<sys:version>1.0-SNAPSHOT</sys:version>
<sys:type>car</sys:type>
</sys:moduleId>
<sys:dependencies>
<!--some dependency on geronimo 2.1.4 here>
</sys:dependencies>
<sys:hidden-classes/>
<sys:non-overridable-classes/>
</sys:environment>
<context-root>/hellocontext</context-root>
</web-app>
I can see following benefits by creating these kind of sample
archetypes:
1, User does not need to install a subversion to get the samples or
to download the samples zip.
He just need to install a maven to get all kinds of samples
whatever he wants.
2, geronimo version could be provided at sample project generating
time.
As I know, most of samples content does not change at all except
those geronimo version update. To keep these "stable" samples to
archetypes can save the time to
maintain every samples per geronimo release. (of course, for those
unstable samples, we still need to update them per geronimo release)
3, Other key properties in geronimo deployment plan could be
provided at sample project generating time.
This is helpful for user to understand the essential elements of
geronimio deployment plan.
Current maven archetype is using velocity template which is very
powerful to generate project contents based on user's input.
It looks to me as if there are no plans in the javaee apps parts of
the samples apps, so the plans are all in the plugin modules. I
really don't think we want to duplicate the functionality of the
existing archetypes that do a fine job of creating projects for java
ee apps. If you really do want to help people make projects like
the sample apps, the way to do that is to improve the plugin
archetype. I like your ideas about what can be generated in the
geronimo plan. I think we can also generate a lot more of the pom
than we do now.
I would be very tempted to try stringtemplate instead of velocity.
I think its clear MVC separation and roots in code generation
technology (it's the output stage for antlr) may prove useful for
the manipulation we need to do.
While enabling people to reproduce our sample apps is probably
better than what we do now, I'd prefer to help them with apps that
actually do a realistic amount of work. I think improving the
plugin archetype would be a really big help.
thanks
david jencks
On Tue, Mar 17, 2009 at 1:41 AM, David Jencks
<[email protected]> wrote:
On Mar 16, 2009, at 12:18 AM, Shawn Jiang wrote:
Maven archetype is a powerful tools to make life easier when
developers want to create a project from scratch. Currently,
Geronimo is using archetype to create new testsuite/assembly/
plugin/ quickly.
1: local -> geronimo-archetype-testsuite (testsuite)
2: local -> geronimo-assembly-archetype (geronimo-assembly-
archetype)
3: local -> geronimo-plugin-archetype (geronimo-plugin-archetype)
4: local -> xmlbeans-maven-plugin (xmlbeans-maven-plugin)
Which can only help geronimo developers/contributors when they
want to create their new testsuite/assembly/plugin. There's no
archetype for geronimo users who want to create their geronimo
Apps/Assets easily from scratch. I'm wondering if geronimo could
provide a set of new archetypes for Apps/assets running in
geronimo. Some initial candidates are:
1, war-geronimo-archetype
A simple war which contains the geronimo deployment plan.
2, ear-geronimo-archtype
A maven project that contains 4 modules:
1, a war module
2, a ear module
3, a ejb module
4, a client module.
3, datasource-geronimo-archtypes
There should be many sub archtypes to create different datasources
plan.
4, security-geronimo-archtypes
There should be many sub archtypes to create different(properties/
db/ldap) realms.
5, GBean-geronimo-archtype
A simple GBean development archtype with two mode: normal/annotation
Any comments ?
I don't quite understand yet what these archetypes would provide.
The most plausible thing would be to add a "fill in the blanks"
plan template. I'm not sure an archetype is the best way to do
this, but I'm certainly open to seeing how it works.
Our existing plugin and assembly archetypes are rather obsolete and
don't use the current archetype infrastructure. I wonder if a more
practical project would be to update especially the plugin
archetype to figure out what kind of thing you are building a
plugin for and include the appropriate plan template and add the
appropriate dependencies and car-maven-plugin configuration.
I guess if we could figure out how to this more or less the same
logic could be used to modify an existing javaee maven project to
insert the appropriate geronimo plan template in the appropriate
directory. I don't think that dupicating the war, ear, etc
archetypes is likely to be worth the effort.
Anyway I'd certainly like to know more about what you have in mind
and am very glad someone else is thinking about maven archetype
support!
thanks
david jencks
--
Shawn
--
Shawn
--
Shawn
