GWT Maven build with maven profiles - gwt

Question

I'm attempting to use capabilities provided by maven profiles to build customized builds for different server environments. What I'm attempting to do is combine maven resource filtering
<resources>
<resource>
<directory>src/main/resources</directory>
<filtering>true</filtering>
</resource>
</resources>
with it's profile mechanism
<profiles>
<profile>
<id>mock</id>
<properties>
<application-url>http://mock-server-url</application-url>
</properties>
<activation>
<activeByDefault>true</activeByDefault>
</activation>
</profile>
</profiles>
to convert this value in a file named server.cfg
${application-url}
to something I can use here:
public interface ServerResource extends ClientBundle {
public static final ServerResource INSTANCE = GWT.create(ServerResource.class);
#Source("server.cfg")
public TextResource server();
}
I can see that the value's been replaced in WEB-INF/classes but it doesn't appear that GWT used the file with the replacement to create the application javascript. How can I do this?

Answer 1

Using GWT compiler permutations to apply this kind of configuration is in my opinion a very bad idea. One of the most common complaints about GWT is the time it takes to compile, and by doing this you're just adding to the problem.
Configuration should usually be read from configuration files (surprise!), like shown here.
Anyway, what you're trying to do seems to me impossible. You cannot tell the client-side code to which server it should connect. This would violate the same-origin policy! The app can only communicate with the server it came from.
To have different apps running in different URLs, you would need to deploy several GWT apps with different names (even if they are basically the same). Then, you would just have to type the correct URL for each app (version) in the browser, and it will "look" at the right app. So you could have URLs like this:
http://myserver.com/app1
http://myserver.com/app2
In order to make a request to a different app running in the same server as the GWT application, you can do something like this:
String serviceUrl = "/app2/someService"; // or some other String sourced from a config file, using a GWT ClientResource for example
RequestBuilder rb = new RequestBuilder(RequestBuilder.GET,
serviceUrl);
try {
// send request from app1 to app2
rb.sendRequest(null, new RequestCallback() {
#Override
public void onResponseReceived(Request request,
Response response) {
log.info("Response: " + response.getStatusText());
// if response is 200 it's ok, you can read the outputStream to see what's in there
}
#Override
public void onError(Request request, Throwable exception) {
log.warning("Request Error", exception);
// do something more
}
});
} catch (RequestException e) {
log.warning("Request Exception", e);
// getting here means trouble with the connection or service!
}

Answer 2

I solved what I was trying to accomplish without the use of maven profiles or the GWT ClientBundle (which I never did get to work in the way I had intended when I wrote the question).
Here were the main issues I hoped to solve using maven profiles and the workaround I employed to solve the issue at hand:
Use Mock MVP Models in Hosted Mode
// inside the initialization for my model locator
boolean hostedMode = GWT.getPermutationStrongName().equals("HostedMode");
if (hostedMode) {
// instantiate mock models
} else {
// instantiate real models to call REST web services
}
Provide real models with correct RESTful server URL
I was able to accomplish this because my GWT app and the RESTful web service url follow a set naming convention. I basically strip the trailing '/' from the URL and append '_services"
String createServicesBaseUrl() {
StringBuffer baseUrl = new StringBuffer(GWT.getHostPageBaseURL());
int length = baseUrl.length();
baseUrl.replace(length-1, length, "_services");
return baseUrl.toString();
}
Enable testing of as much of the MVP Presenter (Activities & Places) as I could
I was already injecting the model locator into my Activity classes, so replacing that with a mock model locator for use by JUnit was straightforward. I did the same for my views and abstracted away some of the other code which didn't seem to work outside of the browser (like the GWT PlaceController).
All in all my build is much the same, but I learned how to gain a lot of flexibility in testing, configuring the server instance my GWT application connects with, and which model my application uses (dependent on hosted vs server mode).