Determining runtime environment on Google App Engine

Google recommends to determine the runtime environment of a Google App Engine application using ServletContext.getServerInfo(). In development this will be "Google App Engine Development/x.x.x" and in production it will be "Google App Engine/x.x.x".

But in some places, like Service classes, you might not have a access to the ServletContext object. Easiest way to work around this is to create a ServletContextListener which determines the environment and then stores it in a System property. The System property can be accesses from any place. You can of course also write a (static) Environment.get() utility class which return the System property.

Below you can find an example implementation of such a ServletContextListener to determine the runtime environment. As you can read the javadocs you can also use other environments then development and production. To start the application in test mode, just add a VM argument when you start your application using "-Druntime.environment=test".

import javax.servlet.ServletContextEvent;
import javax.servlet.ServletContextListener;

/**
 * Simple listener to set the App Engine runtime environment via a system property.
 *
 * Example web.xnl definition:
 *
 * <listener>
 *     <listener-class>com.footdex.web.listener.RuntimeEnvironmentConfigurerListener</listener-class>
 * </listener>
 *
 * The runtime environment can then be accessed from any place (without reference to
 * ServletContext) via System.getProperty(RuntimeEnvironmentConfigurerListener.KEY).
 *
 * In case a system property is already set for the RuntimeEnvironmentConfigurerListener.KEY,
 * the system property will not be overridden. This allows to startup the application
 * in a custom environment other than development or production. This can be done typically
 * using a VM argument like "-Druntime.environment=test". 
 *
 * @author Marcel Overdijk
 */
public class RuntimeEnvironmentConfigurerListener implements ServletContextListener {

    public static final String KEY = "runtime.environment";
    public static final String DEVELOPMENT = "development";
    public static final String PRODUCTION = "production";

    @Override
    public void contextInitialized(ServletContextEvent event) {
        if (System.getProperty(KEY) == null) {
            String serverInfo = event.getServletContext().getServerInfo();
            // ServletContext.getServerInfo() will return "Google App Engine Development/x.x.x" 
            // when run locally, and "Google App Engine/x.x.x" otherwise.
            if (serverInfo.contains("Development")) {
                System.setProperty(KEY, DEVELOPMENT);
            }
            else {
                System.setProperty(KEY, PRODUCTION);
            }
        }
    }

    @Override
    public void contextDestroyed(ServletContextEvent event) {
    }
}

Google App Engine Datastore doubts

I have been looking into Google App Engine since the beginning. I played a little bit with it using Python and Django but never did some serious work. As GAE/J was released back in April I decided to have a serious look by building a real application.

I mean the GAE promise is everything I need: easy to build, easy to maintain, easy to scale and I can run it on Google's infrastructure. And when my application won't be the next Twitter, most likely it will cost me nothing to host it as Google offers enough free quotas for medium and small application. I should also mention that now Grails also runs on GAE (using the App Engine plugin) my life couldn't get easier, but...

The only thing GAE does not offer is a relational datastore. GAE's Datastore, based on BigTable, is a hierarchical, schema-less, non-relational datastore.

One of the nice things relational databases support are the aggregate functions like count, sum, avg etc. If you are using GAE you must "Shift processing from reads to writes" as mentioned in The Softer Side Of Schemas - Mapping Java Persistence Standards To the Google App Engine Datastore presented at Google I/O. In practice this means that when inserting or updating a record you must calculate and store the count, sum, avg, etc. values of the whole table. But what if you have many conditional groupings? Let's take an example:

@Entity
class Person {
    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    String id;

    String name;
    //Date birthdate;
    Integer age;
    String address;
    String city;
    String postalCode;
}

Now imagine I want to be able to (1) get the total number of persons and (2) the average age. Everytime I would create, update or delete a person I should calculate both and store them. Not a big deal, but in my opinion already to much hassle.

But then I also want these values per city... So for each city I should store these values. Next I want them also per postalCode or a specific range of the postalCode. You see where I'm going to? Every create, update or delete of a person would result in a lot of caluculation and storing values.

And then worse, my agile product owner comes in and want to add a boolean indicating a person is a male or not. And guess what, number of persons and average age should additionally be calculated also for this male/female indicator, and to make it worse also in combination with city and postalCode...

I truly believe GAE's Datastore is suitable for many applications don't needing features part of most relational databases. But if you need aggregate queries or common SQL functions (both most likely needed when you are trying to get some analytical data) you should really think twice.

I hope one day GAE's current Datastore will support aggregate functions or additionally a relational datastore will be provided. Then I think GAE would be my preferred Cloud platform.