On 03/14/2013 12:22 PM, Geeb Otron wrote:
Hello,
My question is about auto-generation of CXF code in test and production
environments.
My web app:
1. Consumes a remote web-service
2. Connects to an oracle DB
3. Uses the Spring framework
4. Uses Maven
5. Runs on tomcat and is exported as a WAR file.
I have different environments in which I need to run my app, and ultimately
publish into a production environment, using a production instance of
Oracle, and of the web-service.
What I want to know is the best-practice in terms of having the minimum
amount of environmental config inside my app and the maximum in the
environment in which the app runs.
For example, with the Oracle database, a connection pool (C3P0) is
configured and managed by Tomcat, so my app simply uses a named resource to
grab a database connection object (configured as the web-app starts in the
Spring XML config files, and in the web.xml file)
So irrespective of where the web-app is running the Tomcat instance
provides the connection pool for the the correct database: No configuration
is required in my app other than getting the <resource-ref> section of the
web.xml file correct in the first place, as all instances of Tomcat uses
the same name for their DB connection resource.
Is there a similar approach for configuring the reference to the service
endpoint on the instance of Tomcat that hosts my web-app?
I use the Maven cxf-codegen-plugin and currently have to regenerate and
rebuild the underlying CXF source code as I move from consuming the test
version of the web-service, to the production one.
Make sure you use a WSDL location similar to
<wsdlLocation>classpath:DoubleIt.wsdl</wsdlLocation>
(i.e., use classpath: instead of a hardcoded string), so the generated
artifacts will not be pointing to a specific folder location on your
machine but the supplied classpath instead.
Ideally I would like to generate the CXF code once (as the service
endpoints use pretty much the same public interfaces), but for the app to
pick up the specific URL of the service from the Tomcat environment and it
not have to be hard-coded into the web-app.
CXF largely already does that, it follows this algorithm when it
replaces the URL in the WSDL you provide it:
http://www.jroller.com/gmazza/entry/web_service_tutorial#notes (note #2)
(I'm assuming you're talking about hosting a web service provider on
Tomcat and not just a client, in the latter case it will rely on the
hardcode URL in the WSDL.)
Currently I use a property inside the Maven pom.xml to replace the local
path the WSDL file as I generate code for a specific environment. (I keep
local copies of the WSDL files for all the instances of the web-service)
As I say, ideally I would like to bypass having to regenerate and compile
the code completely if possible, as it seems a faulty approach to tamper
with an artifact that has been signed-off in testing, before it reaches
production!
Thanks for any advice!
