Question

Currently, we use a giant configuration object that is serialized to/from XML. This has worked fine for the most part, but we are finding that in the case of power loss and application crashes, that the file could be left in a state that renders it unable to deserialize properly, effectively corrupting the configuration information.

I would like to use the built-in app.config, but it doesn’t seem to easily support custom classes. For example, with XML serialization, I can easily serialize a generic list<ComplexClass> with no extra code. It just works. It seems that when using app.config, you have to provide a ton of information and custom classes for this to work. Plus, most of the “custom configuration” tutorials are from circa-2007 and may be outdated for all I know. Does anyone have information to the latest way to do this in .NET 4.0?

In addition, when a problem occurs in the application, 9/10 times it is because of an improper configuration. App.config likes to store user-changeable settings in a very inaccessible location for users who aren’t familiar with hidden directories and such. Is there any way to have a single location to store the config file, which the user can easily email us when a problem occurs?

Or, is all this easier than I remember it being in early 2.0 days? Any links or quick examples of how to easily do custom app.config information would be great.

As a further example, this is a pared-down version of one of the types of objects I want to serialize as List<Alarm>, as the amount of Alarms can vary or be empty. Is there an analogous way to store something like this in app.config?

[Serializable]
public class Alarm
{
    [Serializable]
    public class AlarmSetting
    {
        public enum AlarmVariables { Concentration, RSquared }
        public enum AlarmComparisons { LessThan, GreaterThan }

        [Description("Which entity is being alarmed on.")]
        public AlarmVariables Variable { get; set; }
        [Description("Method of comparing the entity to the setpoint.")]
        public AlarmComparisons Comparator { get; set; }
        [Description("Value at which to alarm.")]
        public Double Setpoint { get; set; }
    }

    public String Name { get; set; }
    public Boolean Enabled { get; set; }
    public String Parameter { get; set; }
    public List<AlarmSetting> AlarmSettings { get; set; }
    public System.Drawing.Color RowColor { get; set; }
}

Answer 1

I would suggest moving away from any sort of config file and instead use some type of local database such as sqlite or sql server express which is much more resilient to app crashes.

IMHO, config settings shouldn’t be a default container for user settings. To me a config file is there to make sure the app runs in the given environment. For example, defining connection strings or polling rates or things of that nature.

User settings, especially ones that change often, need a better storage mechanism such as a local database. Unless, of course, it’s a client/server application. In which case those settings should be up at the server itself and only persisted locally if the app has to work in a disconnected state.

The example you gave, one of configuring what appears to be one or more alarms is a perfect example of something that belongs in a database table.