HashSet<string> with StringComparer.InvariantCultureIgnoreCase

[David Boike]

I need to have a model with a case-insensitive hash set.

public class MyModel
{
     public MyModel()
     {
          Set = new HashSet<string>(StringComparer.InvariantCultureIgnoreCase);
     }

     public HashSet<string> Set { get; set; }
}

And this is stored in JSON the same as an array or any other enumerable would be:

{
  "Set": [
    "Item0",
    "Item1",
  ]
}

The problem comes when I load the model back from the database. The HashSet is constructed with the wrong comparer, and eventually you might get this:

{
  "Set": [
    "Item0",
    "ITEM0",
    "item0",
    "iTeM0",
    "Item1",
  ]
}

The only thing I can think of is to create a class that forces the comparer:

public class StringSetIgnoreCase : HashSet<string>
{
    public StringSetIgnoreCase()
        : base(StringComparer.InvariantCultureIgnoreCase)
    {
    }
}

Then I would use StringSetIgnoreCase as the type in my model instead of HashSet<string>.

But of course this has some problems:

• Only handles sets of strings
• And only the one type of string comparer
• Requires other model developers in my organization to know about this issue and this class - there isn't any way for me to hide it in the architecture for their own good.

Is there any better way to do this that addresses those concerns?

[Matt Warren]

You can change the default behaviour by making a CustomerJsonSerializer and registering it with the doc store, there's a nice example here https://groups.google.com/d/msg/ravendb/e6ScMqPwpBs/Ib1n6poes1oJ. It not exactly the same scenario as yours, but should give you an idea of the basics involved.

[Oren Eini (Ayende Rahien)]

You can make it work with a Document Conversion Listener, or by modifying the serializer using CustomizeJsonSerializer