Many Web.config have merged together

You can have more than one Web.config in an application and they will all be merged in the end, which is fine. Also, if you have a problem with the configuration for a given project it will usually show up in your development environment right away and will not make it to production.

But here is a problem that occurred to me a few times and that had me pretty stumped at first. When deploying an ASP.NET MVC or ASP.NET WebForms application to an IIS (Internet Information Service) server, it looked like configurations from unrelated Web.config or other .config files merged together. Errors were popping up from assembly versions or configurations that should not even be used in the new application! So what were those mystery configurations?

In IIS, the Sites folder is a container for all the applications and virtual directories of the server. The Sites folder has a physical root directory even it if doesn’t contain files specific to an application. In fact, configurations files from that folder are merged with the configuration of all your applications. Here is the configuration in IIS for the physical path of that folder:

DefaultWebSiteIn my case, the root folder pointed to another application on the same server, so the configurations made no sense for the new application. To solve the problem, I just made sure than the wwwroot folder was empty or contained only configurations that I intended to share with all the applications hosted on the server.

For references purpose, here is the configuration for the physical path of each individual application. You don’t need to point to a folder under wwwroot as long as the folder used has the proper permissions to be seen from the web.


Storing the state of an activity of your Android application

This is the last post in my series about saving data in your Android application. The previous posts went over the various way to save data in your application:

Introduction : How to save data in your Android application
Saving data to a file in your Android application
Saving preferences in your Android application
Saving to a SQLite database in your Android application

This final post will explain when you should save the current state of your application so you users do not lose their data. There are two types of states that can be saved: you can save the current state of the UI if the user is interrupted while entering data so he can resume input when the application is started again, or you can save the data to a more permanent store of data to access it at any time.

Saving the current UI state

When another application is launched and hides your application, the data your user entered may not be ready to be saved to a permanent store of data yet since the input is not done. On the other hand, you should still save the current state of the activity so the user don’t lose all their work if a phone call comes in for example. A configuration change with the Android device like rotating the screen will also have the same effect, so this is another good reason to save the state.

When one of those event or another event that requires saving the state of the activity occurs, the Android SDK calls the onSaveInstanceState method of the current activity, which receives an android.os.Bundle object as a parameter. If you use standard views from the Android SDK and those views have unique identifiers, the state of those controls will be automatically saved to the bundle. But if you use multiple instances of a view that have the same identifier, for example by repeating a view using a ListView, the values entered in your controls will not be saved since the identifier is duplicated. Also, if you create your own custom controls, you will have to save the state of those controls.

If you need to manually save the state, you must override the onSaveInstanceState method and add your own information to the android.os.Bundle received as a parameter with pairs of key/values. This information will then be available later on when the activity needs to be restored in its onCreate and onRestoreInstanceState methods. All primitives types or arrays of values of those types can be saved to a bundle. If you want to save objects or an array of objects to the bundle they must implement the or android.os.Parcelable interfaces.

To demonstrate saving to a state, I will use an upgraded version of the application used in the article about saving to a database that is available on GitHub at The application manages row counters used for knitting project, but it had no way to create a project. In the new version, the user can now create a new project,  and the state of the project being created needs to be saved if the user is interrupted while creating the project. For demonstration purpose, numerical values are entered using a custom CounterView control that does no handle saving the state, so we must save the state of each counter manually to the bundle.

public void onSaveInstanceState(Bundle savedInstanceState) {
   CounterView rowCounterAmountView = (CounterView)this.findViewById(;
   savedInstanceState.putInt(ROW_COUNTERS_AMOUNT_STATE, rowCounterAmountView.getValue());

   CounterView rowAmountView = (CounterView)this.findViewById(;
   savedInstanceState.putInt(ROWS_AMOUNT_STATE, rowAmountView.getValue());

   // Call the superclass to save the state of all the other controls in the view hierarchy

When the user navigates back to the application, the activity is recreated automatically by the Android SDK from the information that was saved in the bundle. At that point, you must also restore the UI state for the your custom controls. You can restore the UI state using the data you saved to the bundle from two methods of your activity: the onCreate method which is called first when the activity is recreated or the onRestoreInstanceState method that is called after the onStart method. You can restore the state in one method or the other and it most cases it won’t matter, but both are available in case some initialization needs to be done after the onCreate and onStart methods. Here are the two possible ways to restore the state from the activity using the bundle saved in the previous example :

protected void onCreate(Bundle savedInstanceState) {
   [...Normal initialization of the activity...]

   // Check if a previously destroyed activity is being recreated.
   // If a new activity is created, the savedInstanceState will be empty
   if (savedInstanceState != null) {
      // Restore value of counters from saved state
      CounterView rowCounterAmountView;
      rowCounterAmountView  = (CounterView)this.findViewById(;

      CounterView rowAmountView = (CounterView)this.findViewById(;
protected void onRestoreInstanceState(Bundle savedInstanceState) {
   // Call the superclass to restore the state of all the other controls in the view hierarchy

   // Restore value of counters from saved stat
   CounterView rowCounterAmountView = (CounterView)this.findViewById(;

   CounterView rowAmountView = (CounterView)this.findViewById(;

Remember, saving data in a bundle is not meant to be a permanent store of data since it only stores the current state of the view: it is not part of the activity lifecyle and is only called when the activity needs to be recreated or it is sent to the background. This means that the onSaveInstanceState method is not called when the application is destroyed since the activity state could never be restored. To save data that should never be lost you should save the data to one of the permanent data store described earlier in this series. But when should this data be stored?

Saving your data to a permanent data store

If you need to save data to a permanent data store when the activity is send to the background or destroyed for any reason, your must save your data in the onPause method of the activity. The onStop method is called only if the UI is completely hidden, so you cannot rely on it being raised all the time. All the essential data must be saved at this point because you have no control on what happens after: the user may kill the application for example and the data would be lost.

In the previous version of the application, when the user incremented the counter for a row in a project the application saved the current value of the counter to the database every time. Now we’ll save the data only when the user leaves the activity, and saving at each counter press is no longer required:

public void onPause() {
   ProjectsDatabaseHelper database = new ProjectsDatabaseHelper(this);
   // Update the value of all the counters of the project in the database since the activity is
   // destroyed or send to the background
   for (RowCounter rowCounter: mRowCounters) {

Later on, if your application was not destroyed and the user accesses the activity again there are two possible processes that can occurs depending on how the Android OS handled your activity. If the activity was still in memory, for example if the user opened another application and came back immediately, the onRestart method is first called, followed by a call to the onStart method and finally the OnResume method is called and the activity is shown to the user. But if the activity was recycled and is being recreated,  for example if the user rotated the device so the layout is recreated, the process is the same as the one for a new activity: the onCreate method is first called, followed by a call to the onStart method and finally the onResume method is called and the activity is shown to the user.

So, if you want to use the data that was saved to a permanent data store to initialize controls in your activity that lose their state, you should put your code in the onResume method since it is always called, regardless of if the activity was recreated or not. In the previous example, it is not necessary to explicitly restore the data since no custom controls were used: if the activity was recycled, it is recreated from scratch and the onCreate method initialize the controls from the data in the database. If the activity is still in memory, there is nothing else to do: the Android SDK handles showing the values as they were first displayed as explained earlier in the section about saving UI states. Here is a reminder of what happens in the onCreate method:

protected void onCreate(Bundle savedInstanceState) {
   Intent intent = getIntent();
   long projectId = intent.getLongExtra("project_id", -1);
   // Gets the database helper to access the database for the application
   ProjectsDatabaseHelper database = new ProjectsDatabaseHelper(this);
   // Use the helper to get the current project
   Project currentProject = database.getProject(projectId);

   TextView projectNameView = (TextView)findViewById(; 
		// Initialize the listview to show the row counters for the project from 
		// the database
   ListView rowCounterList = (ListView)findViewById(;
   mRowCounters = currentProject.getRowCounters();
   ListAdapter rowCounterListAdapter = new RowCounterAdapter(this,

This concludes the series about saving data in your Android application. You now know about the various types of data storage that are available for Android applications and when should they be used so your users never lose their data and have the best user experience possible.

Building maintainable Android applications

Why should you care about maintainability when everything already works? It’s not a bad thing to try out new things with prototypes to increase your knowledge, but when you start building applications that will be used for many years and that will grow larger and more complex, you want to make sure you can keep adding to it without running into problems. When many people need your software to do their work every day, you cannot afford to build something that will collapse under it’s own weight. So, here are a few guidelines to help you:

Keep your code clean and documented
When you have to go back regularly to modify your old code, having clear names and documentation is very valuable. You only write a piece of code one, but it will be read again many times so you should make that job as easy as possible. Having a consistent coding standard also helps a lot with readability: it is jarring to read code that jumps from one coding style to another, and you want to concentrate on the code and not how it looks.

Also, I don’t believe that all code can be self-documenting code. It’s a good thing to aim for, but sometimes you have to do hacks that are not as clean: a little explanation can go a long way when you try understanding it again a few years down the road. In other cases, for example when you implement a complex bit of business logic, you will need to know what the code was supposed to do if it does not behave as expected or if the logic needs to be changed later on.

Don’t do complicated when you can do simple
In keeping with good Agile principles, you don’t want to plan everything that’s even going to happen in you application before getting started. Don’t get me wrong, you should have an idea of where you’re going and have a basic architecture planned out, but keep thing as simple as possible while still being organized. You can always refactor a complex bit of code to add one more layer of abstraction as needed and optimize the bits that are not as effective, but having a relatively simple architecture planned out helps keep everything clean and makes it easy to introduce new team members to a project.

You should also avoid clutter in your code base. First, fix as many compilation warnings as you can: some may not be important to you, but you don’t want to risk missing a new warning that shows a real problem. It’s something small and silly, but it makes it harder every time you need to modify something. Also, only implement the methods you need and don’t keep old code lying around to limit the amount of code you have to maintain. If you need to reuse the code you deleted later on, you can always retrieve it from your source control system.

Limit the number of external libraries
There are many good libraries available in the Android and Java worlds. But while using a library can save you time, it also has has a real cost: when you upgrade your application to support the newest Android version, you also need to upgrade all the libraries you use. Each dependency is a part of your project that can break and on which you have no direct control. Libraries can also stop being upgraded, which will eventually require a re-factoring to an alternative or an effort from your part to maintain the old library, which can be be costly.

On the other hand, using a library it can be worth it if you need to make a quick prototype or to add a component that would take too long to built. But for mission-critical stuff that can make or break your application, you may be better off making your own implementation based on your needs. Fortunately, nothing is stopping you from learning from the existing libraries to make your own thing. You may also prefer to use methods from the standard Android SDK version even if their implementation is not the best: it is the most likely to have been tested on a large range of devices and to be maintained for a long time.

Your product is not your debug version
If your code works when loaded from your development machine to your testing device that’s great but it’s not the end of your job. You need to test your release version on at least a few more device and Android version to be sure it will work when released to the store, since not everyone that will use your application has the same device and OS version than you. You also need to keep testing when new versions of the Android OS keep coming out so it does not break for your users that like to keep their device updated with the latest and greatest.

Also, if your application is not deployed via the Google Play store, you may also wish to add some form of debugging to your application, for example a logging mode that can be toggled on to report a problem to you: it can be pretty hard to help a user when the only information you have is that your application crashed. Finally, you also need to have some documentation ready for your users, unless your application is very simple. A little FAQ on your website may be enough to do the job, but don’t leave your user totally in the dark trying to understand your application. This will reduce the number of support requests and give you a few answers ready to go when the questions start coming in.