Json.NET contract resolver that uses Ruby-style lowercase with underscore naming conventions.

SnakeCaseContractResolver.cs

using Newtonsoft.Json.Serialization;

namespace ConsoleApplication3
{
    public class SnakeCaseContractResolver : DefaultContractResolver
    {
        protected override string ResolvePropertyName(string propertyName)
        {
            return GetSnakeCase(propertyName);
        }

        private string GetSnakeCase(string input)
        {
            if (string.IsNullOrEmpty(input))
                return input;

            var buffer = "";

            for (var i = 0; i < input.Length; i++)
            {
                var isLast = (i == input.Length - 1);
                var isSecondFromLast = (i == input.Length - 2);

                var curr = input[i];
                var next = !isLast ? input[i + 1] : '\0';
                var afterNext = !isSecondFromLast && !isLast ? input[i + 2] : '\0';

                buffer += char.ToLower(curr);

                if (!char.IsDigit(curr) && char.IsUpper(next))
                {
                    if (char.IsUpper(curr))
                    {
                        if (!isLast && !isSecondFromLast && !char.IsUpper(afterNext))
                            buffer += "_";
                    }
                    else
                        buffer += "_";
                }

                if (!char.IsDigit(curr) && char.IsDigit(next))
                    buffer += "_";
                if (char.IsDigit(curr) && !char.IsDigit(next) && !isLast)
                    buffer += "_";
            }

            return buffer;
        }
    }
}

Nice! Thank you!

Thanks!

Hey guys,
This doesn't work if some of it is already PascalCase. anyone come across this issue?
i.e.

 {
    "some_key": 0,
    "SomeOtherKey": 0,
    "some_id": "1",
    "SomeOtherId": "2"
  }

only the first and the third one work

@crallen?

// (Preceded by a lowercase character or digit) (a capital) => The character prefixed with an underscore
var result = Regex.Replace(input, "(?<=[a-z0-9])[A-Z]", m => "_" + m.Value);
result = result.ToLowerInvariant();

• This works for both PascalCase and camelCase.
• It creates no leading or trailing underscores.
• It leaves in tact any sequences of non-word characters and underscores in the string, because they would seem intentional, e.g. __HiThere_Guys becomes __hi_there_guys.
• Digit suffixes are (intentionally) considered part of the word, e.g. NewVersion3 becomes new_version3.
• Digit prefixes follow the original casing, e.g. 3VersionsHere becomes 3_versions_here, but 3rdVersion becomes 3rd_version.

I agree, I was using Regex.Replace(input, @"([a-z0-9])([A-Z])", "$1_$2").ToLowerInvariant();
I think the ?<= in the regex adds unnecessary complexity, am I wrong?

Interesting idea. I'm not sure how the ?<= (lookbehind) is handled by the regex engine. It's that vs. replacing more data, which at least is guaranteed to be linear time. And yours looks simpler too, so agreed.

Works! Thanks.

Nowadays you don't need a custom contract resolver, just do:

var settings = new JsonSerializerSettings
{
    ContractResolver = new DefaultContractResolver
    {
        NamingStrategy = new SnakeCaseNamingStrategy()
    }
};

Hi Chris,
Thank you for your code! :)

Can you please tell me what is the license for the above code?

Regards,
Andrei