We Keep Coding

How can I convert part of an api response from an object to a list? [duplicate]

I'm trying to fix my SendGridPlus library to deal with SendGrid events, but I'm having some trouble with the inconsistent treatment of categories in the API.
In the following example payload taken from the SendGrid API reference, you'll notice that the category property for each item can either be a single string or an array of strings.
[
{
"email": "john.doe#sendgrid.com",
"timestamp": 1337966815,
"category": [
"newuser",
"transactional"
],
"event": "open"
},
{
"email": "jane.doe#sendgrid.com",
"timestamp": 1337966815,
"category": "olduser",
"event": "open"
}
]
It seems my options to make JSON.NET like this are fixing the string before it comes in, or configuring JSON.NET to accept the incorrect data. I'd rather not do any string parsing if I can get away with it.
Is there any other way I can handle this using Json.Net?

The best way to handle this situation is to use a custom JsonConverter.
Before we get to the converter, we'll need to define a class to deserialize the data into. For the Categories property that can vary between a single item and an array, define it as a List<string> and mark it with a [JsonConverter] attribute so that JSON.Net will know to use the custom converter for that property. I would also recommend using [JsonProperty] attributes so that the member properties can be given meaningful names independent of what is defined in the JSON.
class Item
{
[JsonProperty("email")]
public string Email { get; set; }
[JsonProperty("timestamp")]
public int Timestamp { get; set; }
[JsonProperty("event")]
public string Event { get; set; }
[JsonProperty("category")]
[JsonConverter(typeof(SingleOrArrayConverter<string>))]
public List<string> Categories { get; set; }
}
Here is how I would implement the converter. Notice I've made the converter generic so that it can be used with strings or other types of objects as needed.
class SingleOrArrayConverter<T> : JsonConverter
{
public override bool CanConvert(Type objectType)
{
return (objectType == typeof(List<T>));
}
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
JToken token = JToken.Load(reader);
if (token.Type == JTokenType.Array)
{
return token.ToObject<List<T>>();
}
return new List<T> { token.ToObject<T>() };
}
public override bool CanWrite
{
get { return false; }
}
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}
Here is an short program demonstrating the converter in action with your sample data:
class Program
{
static void Main(string[] args)
{
string json = #"
[
{
""email"": ""john.doe#sendgrid.com"",
""timestamp"": 1337966815,
""category"": [
""newuser"",
""transactional""
],
""event"": ""open""
},
{
""email"": ""jane.doe#sendgrid.com"",
""timestamp"": 1337966815,
""category"": ""olduser"",
""event"": ""open""
}
]";
List<Item> list = JsonConvert.DeserializeObject<List<Item>>(json);
foreach (Item obj in list)
{
Console.WriteLine("email: " + obj.Email);
Console.WriteLine("timestamp: " + obj.Timestamp);
Console.WriteLine("event: " + obj.Event);
Console.WriteLine("categories: " + string.Join(", ", obj.Categories));
Console.WriteLine();
}
}
}
And finally, here is the output of the above:
email: john.doe#sendgrid.com
timestamp: 1337966815
event: open
categories: newuser, transactional
email: jane.doe#sendgrid.com
timestamp: 1337966815
event: open
categories: olduser
Fiddle: https://dotnetfiddle.net/lERrmu
EDIT
If you need to go the other way, i.e. serialize, while keeping the same format, you can implement the WriteJson() method of the converter as shown below. (Be sure to remove the CanWrite override or change it to return true, or else WriteJson() will never be called.)
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
List<T> list = (List<T>)value;
if (list.Count == 1)
{
value = list[0];
}
serializer.Serialize(writer, value);
}
Fiddle: https://dotnetfiddle.net/XG3eRy

I was working on this for ages, and thanks to Brian for his answer.
All I am adding is the vb.net answer!:
Public Class SingleValueArrayConverter(Of T)
sometimes-array-and-sometimes-object
Inherits JsonConverter
Public Overrides Sub WriteJson(writer As JsonWriter, value As Object, serializer As JsonSerializer)
Throw New NotImplementedException()
End Sub
Public Overrides Function ReadJson(reader As JsonReader, objectType As Type, existingValue As Object, serializer As JsonSerializer) As Object
Dim retVal As Object = New [Object]()
If reader.TokenType = JsonToken.StartObject Then
Dim instance As T = DirectCast(serializer.Deserialize(reader, GetType(T)), T)
retVal = New List(Of T)() From { _
instance _
}
ElseIf reader.TokenType = JsonToken.StartArray Then
retVal = serializer.Deserialize(reader, objectType)
End If
Return retVal
End Function
Public Overrides Function CanConvert(objectType As Type) As Boolean
Return False
End Function
End Class
then in your class:
<JsonProperty(PropertyName:="JsonName)> _
<JsonConverter(GetType(SingleValueArrayConverter(Of YourObject)))> _
Public Property YourLocalName As List(Of YourObject)
Hope this saves you some time

As a minor variation to the great answer by Brian Rogers, here are two tweaked versions of SingleOrArrayConverter<T>.
Firstly, here is a version that works for all List<T> for every type T that is not itself a collection:
public class SingleOrArrayListConverter : JsonConverter
{
// Adapted from this answer https://stackoverflow.com/a/18997172
// to https://stackoverflow.com/questions/18994685/how-to-handle-both-a-single-item-and-an-array-for-the-same-property-using-json-n
// by Brian Rogers https://stackoverflow.com/users/10263/brian-rogers
readonly bool canWrite;
readonly IContractResolver resolver;
public SingleOrArrayListConverter() : this(false) { }
public SingleOrArrayListConverter(bool canWrite) : this(canWrite, null) { }
public SingleOrArrayListConverter(bool canWrite, IContractResolver resolver)
{
this.canWrite = canWrite;
// Use the global default resolver if none is passed in.
this.resolver = resolver ?? new JsonSerializer().ContractResolver;
}
static bool CanConvert(Type objectType, IContractResolver resolver)
{
Type itemType;
JsonArrayContract contract;
return CanConvert(objectType, resolver, out itemType, out contract);
}
static bool CanConvert(Type objectType, IContractResolver resolver, out Type itemType, out JsonArrayContract contract)
{
if ((itemType = objectType.GetListItemType()) == null)
{
itemType = null;
contract = null;
return false;
}
// Ensure that [JsonObject] is not applied to the type.
if ((contract = resolver.ResolveContract(objectType) as JsonArrayContract) == null)
return false;
var itemContract = resolver.ResolveContract(itemType);
// Not implemented for jagged arrays.
if (itemContract is JsonArrayContract)
return false;
return true;
}
public override bool CanConvert(Type objectType) { return CanConvert(objectType, resolver); }
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
Type itemType;
JsonArrayContract contract;
if (!CanConvert(objectType, serializer.ContractResolver, out itemType, out contract))
throw new JsonSerializationException(string.Format("Invalid type for {0}: {1}", GetType(), objectType));
if (reader.MoveToContent().TokenType == JsonToken.Null)
return null;
var list = (IList)(existingValue ?? contract.DefaultCreator());
if (reader.TokenType == JsonToken.StartArray)
serializer.Populate(reader, list);
else
// Here we take advantage of the fact that List<T> implements IList to avoid having to use reflection to call the generic Add<T> method.
list.Add(serializer.Deserialize(reader, itemType));
return list;
}
public override bool CanWrite { get { return canWrite; } }
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
var list = value as ICollection;
if (list == null)
throw new JsonSerializationException(string.Format("Invalid type for {0}: {1}", GetType(), value.GetType()));
// Here we take advantage of the fact that List<T> implements IList to avoid having to use reflection to call the generic Count method.
if (list.Count == 1)
{
foreach (var item in list)
{
serializer.Serialize(writer, item);
break;
}
}
else
{
writer.WriteStartArray();
foreach (var item in list)
serializer.Serialize(writer, item);
writer.WriteEndArray();
}
}
}
public static partial class JsonExtensions
{
public static JsonReader MoveToContent(this JsonReader reader)
{
while ((reader.TokenType == JsonToken.Comment || reader.TokenType == JsonToken.None) && reader.Read())
;
return reader;
}
internal static Type GetListItemType(this Type type)
{
// Quick reject for performance
if (type.IsPrimitive || type.IsArray || type == typeof(string))
return null;
while (type != null)
{
if (type.IsGenericType)
{
var genType = type.GetGenericTypeDefinition();
if (genType == typeof(List<>))
return type.GetGenericArguments()[0];
}
type = type.BaseType;
}
return null;
}
}
It can be used as follows:
var settings = new JsonSerializerSettings
{
// Pass true if you want single-item lists to be reserialized as single items
Converters = { new SingleOrArrayListConverter(true) },
};
var list = JsonConvert.DeserializeObject<List<Item>>(json, settings);
Notes:
The converter avoids the need to pre-load the entire JSON value into memory as a JToken hierarchy.
The converter does not apply to lists whose items are also serialized as collections, e.g. List<string []>
The Boolean canWrite argument passed to the constructor controls whether to re-serialize single-element lists as JSON values or as JSON arrays.
The converter's ReadJson() uses the existingValue if pre-allocated so as to support populating of get-only list members.
Secondly, here is a version that works with other generic collections such as ObservableCollection<T>:
public class SingleOrArrayCollectionConverter<TCollection, TItem> : JsonConverter
where TCollection : ICollection<TItem>
{
// Adapted from this answer https://stackoverflow.com/a/18997172
// to https://stackoverflow.com/questions/18994685/how-to-handle-both-a-single-item-and-an-array-for-the-same-property-using-json-n
// by Brian Rogers https://stackoverflow.com/users/10263/brian-rogers
readonly bool canWrite;
public SingleOrArrayCollectionConverter() : this(false) { }
public SingleOrArrayCollectionConverter(bool canWrite) { this.canWrite = canWrite; }
public override bool CanConvert(Type objectType)
{
return typeof(TCollection).IsAssignableFrom(objectType);
}
static void ValidateItemContract(IContractResolver resolver)
{
var itemContract = resolver.ResolveContract(typeof(TItem));
if (itemContract is JsonArrayContract)
throw new JsonSerializationException(string.Format("Item contract type {0} not supported.", itemContract));
}
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
ValidateItemContract(serializer.ContractResolver);
if (reader.MoveToContent().TokenType == JsonToken.Null)
return null;
var list = (ICollection<TItem>)(existingValue ?? serializer.ContractResolver.ResolveContract(objectType).DefaultCreator());
if (reader.TokenType == JsonToken.StartArray)
serializer.Populate(reader, list);
else
list.Add(serializer.Deserialize<TItem>(reader));
return list;
}
public override bool CanWrite { get { return canWrite; } }
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
ValidateItemContract(serializer.ContractResolver);
var list = value as ICollection<TItem>;
if (list == null)
throw new JsonSerializationException(string.Format("Invalid type for {0}: {1}", GetType(), value.GetType()));
if (list.Count == 1)
{
foreach (var item in list)
{
serializer.Serialize(writer, item);
break;
}
}
else
{
writer.WriteStartArray();
foreach (var item in list)
serializer.Serialize(writer, item);
writer.WriteEndArray();
}
}
}
Then, if your model is using, say, an ObservableCollection<T> for some T, you could apply it as follows:
class Item
{
public string Email { get; set; }
public int Timestamp { get; set; }
public string Event { get; set; }
[JsonConverter(typeof(SingleOrArrayCollectionConverter<ObservableCollection<string>, string>))]
public ObservableCollection<string> Category { get; set; }
}
Notes:
In addition to the notes and restrictions for SingleOrArrayListConverter, the TCollection type must be read/write and have a parameterless constructor.
Demo fiddle with basic unit tests here.

To handle this you have to use a custom JsonConverter. But you probably already had that in mind.
You are just looking for a converter that you can use immediately. And this offers more than just a solution for the situation described.
I give an example with the question asked.
How to use my converter:
Place a JsonConverter Attribute above the property. JsonConverter(typeof(SafeCollectionConverter))
public class SendGridEvent
{
[JsonProperty("email")]
public string Email { get; set; }
[JsonProperty("timestamp")]
public long Timestamp { get; set; }
[JsonProperty("category"), JsonConverter(typeof(SafeCollectionConverter))]
public string[] Category { get; set; }
[JsonProperty("event")]
public string Event { get; set; }
}
And this is my converter:
using Newtonsoft.Json;
using Newtonsoft.Json.Linq;
using System;
namespace stackoverflow.question18994685
{
public class SafeCollectionConverter : JsonConverter
{
public override bool CanConvert(Type objectType)
{
return true;
}
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
//This not works for Populate (on existingValue)
return serializer.Deserialize<JToken>(reader).ToObjectCollectionSafe(objectType, serializer);
}
public override bool CanWrite => false;
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}
}
And this converter uses the following class:
using System;
namespace Newtonsoft.Json.Linq
{
public static class SafeJsonConvertExtensions
{
public static object ToObjectCollectionSafe(this JToken jToken, Type objectType)
{
return ToObjectCollectionSafe(jToken, objectType, JsonSerializer.CreateDefault());
}
public static object ToObjectCollectionSafe(this JToken jToken, Type objectType, JsonSerializer jsonSerializer)
{
var expectArray = typeof(System.Collections.IEnumerable).IsAssignableFrom(objectType);
if (jToken is JArray jArray)
{
if (!expectArray)
{
//to object via singel
if (jArray.Count == 0)
return JValue.CreateNull().ToObject(objectType, jsonSerializer);
if (jArray.Count == 1)
return jArray.First.ToObject(objectType, jsonSerializer);
}
}
else if (expectArray)
{
//to object via JArray
return new JArray(jToken).ToObject(objectType, jsonSerializer);
}
return jToken.ToObject(objectType, jsonSerializer);
}
public static T ToObjectCollectionSafe<T>(this JToken jToken)
{
return (T)ToObjectCollectionSafe(jToken, typeof(T));
}
public static T ToObjectCollectionSafe<T>(this JToken jToken, JsonSerializer jsonSerializer)
{
return (T)ToObjectCollectionSafe(jToken, typeof(T), jsonSerializer);
}
}
}
What does it do exactly?
If you place the converter attribute the converter will be used for this property. You can use it on a normal object if you expect a json array with 1 or no result. Or you use it on an IEnumerable where you expect a json object or json array. (Know that an array -object[]- is an IEnumerable)
A disadvantage is that this converter can only be placed above a property because he thinks he can convert everything. And be warned. A string is also an IEnumerable.
And it offers more than an answer to the question:
If you search for something by id you know that you will get an array back with one or no result.
The ToObjectCollectionSafe<TResult>() method can handle that for you.
This is usable for Single Result vs Array using JSON.net
and handle both a single item and an array for the same property
and can convert an array to a single object.
I made this for REST requests on a server with a filter that returned one result in an array but wanted to get the result back as a single object in my code. And also for a OData result response with expanded result with one item in an array.
Have fun with it.

Just wanted to add to #dbc excellent response above on the SingleOrArrayCollectionConverter. I was able to modify it to use with a stream from an HTTP client. Here is a snippet (you will have to set up the requestUrl (string) and the httpClient (using System.Net.Http;).
public async Task<IList<T>> HttpRequest<T>(HttpClient httpClient, string requestedUrl, CancellationToken cancellationToken)
{
using (var request = new HttpRequestMessage(HttpMethod.Get, requestedUrl))
using (var httpResponseMessage = await httpClient.SendAsync(request, HttpCompletionOption.ResponseHeadersRead, cancellationToken))
{
if (httpResponseMessage.IsSuccessStatusCode)
{
using var stream = await httpResponseMessage.Content.ReadAsStreamAsync();
using var streamReader = new StreamReader(stream);
using var jsonTextReader = new JsonTextReader(streamReader );
var settings = new JsonSerializerSettings
{
// Pass true if you want single-item lists to be reserialized as single items
Converters = { new SingleOrArrayCollectionConverter(true) },
};
var jsonSerializer = JsonSerializer.Create(settings);
return jsonSerializer.Deserialize<List<T>>(jsonTextReader);
}
I apologize if there are missing brackets or misspellings, it was not easy to paste code in here.

I had a very similar Problem.
My Json Request was completly unknown for me.
I only knew.
There will be an objectId in it and some anonym key value pairs AND arrays.
I used it for an EAV Model i did:
My JSON Request:
{objectId": 2,
"firstName": "Hans",
"email" :[ "a#b.de","a#c.de"],
"name": "Andre",
"something" :["232","123"]
}
My Class i defined:
[JsonConverter(typeof(AnonyObjectConverter))]
public class AnonymObject
{
public AnonymObject()
{
fields = new Dictionary<string, string>();
list = new List<string>();
}
public string objectid { get; set; }
public Dictionary<string, string> fields { get; set; }
public List<string> list { get; set; }
}
and now that i want to deserialize unknown attributes with its value and arrays in it my Converter looks like that:
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
AnonymObject anonym = existingValue as AnonymObject ?? new AnonymObject();
bool isList = false;
StringBuilder listValues = new StringBuilder();
while (reader.Read())
{
if (reader.TokenType == JsonToken.EndObject) continue;
if (isList)
{
while (reader.TokenType != JsonToken.EndArray)
{
listValues.Append(reader.Value.ToString() + ", ");
reader.Read();
}
anonym.list.Add(listValues.ToString());
isList = false;
continue;
}
var value = reader.Value.ToString();
switch (value.ToLower())
{
case "objectid":
anonym.objectid = reader.ReadAsString();
break;
default:
string val;
reader.Read();
if(reader.TokenType == JsonToken.StartArray)
{
isList = true;
val = "ValueDummyForEAV";
}
else
{
val = reader.Value.ToString();
}
try
{
anonym.fields.Add(value, val);
}
catch(ArgumentException e)
{
throw new ArgumentException("Multiple Attribute found");
}
break;
}
}
return anonym;
}
So now everytime i get an AnonymObject i can iterate through the Dictionary and everytime there is my Flag "ValueDummyForEAV" i switch to the list, read the first line and split the values. After that i delete the first entry from the list and go on with iteration from the Dictionary.
Maybe someone has the same problem and can use this :)
Regards
Andre

You can use a JSONConverterAttribute as found here: http://james.newtonking.com/projects/json/help/
Presuming you have a class that looks like
public class RootObject
{
public string email { get; set; }
public int timestamp { get; set; }
public string smtpid { get; set; }
public string #event { get; set; }
public string category[] { get; set; }
}
You'd decorate the category property as seen here:
[JsonConverter(typeof(SendGridCategoryConverter))]
public string category { get; set; }
public class SendGridCategoryConverter : JsonConverter
{
public override bool CanConvert(Type objectType)
{
return true; // add your own logic
}
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
// do work here to handle returning the array regardless of the number of objects in
}
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
// Left as an exercise to the reader :)
throw new NotImplementedException();
}
}

You don't need any custom converters, in this case I am usually creating a very simple JsonConstructor
public partial class Item
{
// ... all class properties
[JsonConstructor]
public Item(JToken category)
{
if (category.GetType().Name == "JArray")
Category = category.ToObject<List<string>>();
else
Category = new List<string> { category.ToString() };
}
public Item() { }
}
after this you can deserialize your json using common code
List<Item> items = JsonConvert.DeserializeObject<List<Item>>(json);

I found another solution that can handle the category as string or array by using object. This way I don´t need to mess up with the json serializer.
Please give it a look if you have the time and tell me what you think. https://github.com/MarcelloCarreira/sendgrid-csharp-eventwebhook
It´s based on the solution at https://sendgrid.com/blog/tracking-email-using-azure-sendgrid-event-webhook-part-1/ but I also added date conversion from timestamp, upgraded the variables to reflect current SendGrid model (and made categories work).
I also created a handler with basic auth as option. See the ashx files and the examples.
Thank you!

How change object type to another when deserialising with JSON.Net or call default serialiser?

When serialise, I write ClassName to object into _CurrentClassName property. And when read json with JSON.Net library I need to change object to value from this property.
{
"Field1": 0,
"Field2": "34",
"_CurrentClassName": "MyCustomClass"
}
class CustomJsonConverter : JsonConverter
{
...
public override bool CanConvert(Type objectType)
{
return objectType.IsClass;
}
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
var value = existingValue;
if (reader.TokenType == JsonToken.Null)
{
return null;
}
else if (reader.TokenType == JsonToken.StartObject)
{
JObject jObject = JObject.Load(reader);
JToken jToken;
if (jObject.TryGetValue("_CurrentClassName", out jToken))
{
var t = jToken.Value<string>();
Type tt = Type.GetType(objectType.Namespace + "." + t);
value = Activator.CreateInstance(tt);
return value;
}
}
return serializer.Deserialize(reader);
}
...
}

Once the object type has been inferred and the object has been instantiated, you can use JsonSerializer.Populate(jObject.CreateReader()) to populate it.
For instance:
public abstract class CustomJsonConverterBase : JsonConverter
{
protected abstract Type InferType(JToken token, Type objectType, object existingValue);
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
if (reader.TokenType == JsonToken.Null)
return null;
var token = JToken.Load(reader);
var actualType = InferType(token, objectType, existingValue);
if (existingValue == null || existingValue.GetType() != actualType)
{
var contract = serializer.ContractResolver.ResolveContract(actualType);
existingValue = contract.DefaultCreator();
}
using (var subReader = token.CreateReader())
{
// Using "populate" avoids infinite recursion.
serializer.Populate(subReader, existingValue);
}
return existingValue;
}
public override bool CanWrite { get { return false; } }
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}
public class CustomJsonConverter : CustomJsonConverterBase
{
public const string ClassPropertyName = #"_CurrentClassName";
protected override Type InferType(JToken token, Type objectType, object existingValue)
{
if (token is JObject)
{
var typeName = (string)token[ClassPropertyName];
if (typeName != null)
{
var actualType = Type.GetType(objectType.Namespace + "." + typeName);
if (actualType != null)
return actualType;
}
}
return objectType;
}
public override bool CanConvert(Type objectType)
{
throw new NotImplementedException();
}
}
CustomCreationConverter<T> also uses serializer.Populate() to fill in the just-allocated object so this is a standard way to solve this problem.
Note that you are partially replicating Json.NET's built-in TypeNameHandling functionality.

If you're not married to the _CurrentClassName property name or its value syntax you could use Json.Net's built in handling of types.
When serializing or deserializing you can pass in a JsonSerializerSettings object controlling the serialization or deserialization.
On this object you can set a TypeNameHandling property that controls how Json.Net serializes and deserializes the exact type being processed.
Here is a LINQPad example:
void Main()
{
var t = new Test { Key = 42, Value = "Meaning of life" };
var json = JsonConvert.SerializeObject(
t, Newtonsoft.Json.Formatting.Indented,
new JsonSerializerSettings { TypeNameHandling = TypeNameHandling.All });
Console.WriteLine(json);
var obj =JsonConvert.DeserializeObject(json,
new JsonSerializerSettings { TypeNameHandling = TypeNameHandling.All });
Console.WriteLine(obj.GetType().FullName);
}
public class Test
{
public int Key { get; set; }
public string Value { get; set; }
}
Output:
{
"$type": "UserQuery+Test, LINQPadQuery",
"Key": 42,
"Value": "Meaning of life"
}
UserQuery+Test
Here you can see that the type of the object being returned from deserialization is the Test class.

Prevent Json.NET from deserializing arrays of strings as arrays of key-value pairs

JsonConvert.DeserializeObject successfully deserializes ['a','b'] as List<KeyValuePair<string, object>>. I would like it to fail, only succeeding when the input string is like [{'Key':'a','Value':'b'}].
Is there a way to accomplish this?

It appears you may have found a bug in Json.NET's KeyValuePairConverter, namely that it assumes the reader is positioned at the beginning of a JSON object rather than checking and validating that it is. You could report an issue on it if you want.
In the meantime, the following JsonConverter will correctly throw a JsonException for your case:
public class KeyValueConverter : JsonConverter
{
interface IToKeyValuePair
{
object ToKeyValuePair();
}
struct Pair<TKey, TValue> : IToKeyValuePair
{
public TKey Key { get; set; }
public TValue Value { get; set; }
public object ToKeyValuePair()
{
return new KeyValuePair<TKey, TValue>(Key, Value);
}
}
public override bool CanConvert(Type objectType)
{
bool isNullable = (Nullable.GetUnderlyingType(objectType) != null);
Type type = (Nullable.GetUnderlyingType(objectType) ?? objectType);
return type.IsGenericType
&& type.GetGenericTypeDefinition() == typeof(KeyValuePair<,>);
}
public override bool CanWrite { get { return false; } } // Use Json.NET's writer.
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
bool isNullable = (Nullable.GetUnderlyingType(objectType) != null);
Type type = (Nullable.GetUnderlyingType(objectType) ?? objectType);
if (isNullable && reader.TokenType == JsonToken.Null)
return null;
if (type.IsGenericType
&& type.GetGenericTypeDefinition() == typeof(KeyValuePair<,>))
{
var pairType = typeof(Pair<,>).MakeGenericType(type.GetGenericArguments());
var pair = serializer.Deserialize(reader, pairType);
if (pair == null)
return null;
return ((IToKeyValuePair)pair).ToKeyValuePair();
}
else
{
throw new JsonSerializationException("Invalid type: " + objectType);
}
}
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}
Then use it like:
string json = #"['a','b']";
var settings = new JsonSerializerSettings { Converters = new JsonConverter[] { new KeyValueConverter() } };
var list = JsonConvert.DeserializeObject<List<KeyValuePair<string, object>>>(json, settings);
Example fiddle.
Update
To force an error when the JSON contains a non-existent property, use JsonSerializerSettings.MissingMemberHandling = MissingMemberHandling.Error.
var settings = new JsonSerializerSettings
{
MissingMemberHandling = MissingMemberHandling.Error,
Converters = new JsonConverter[] { new KeyValueConverter() },
};

JSON.Net throws StackOverflowException when using [JsonConvert()]

I wrote this simple code to Serialize classes as flatten, but when I use [JsonConverter(typeof(FJson))] annotation, it throws a StackOverflowException. If I call the SerializeObject manually, it works fine.
How can I use JsonConvert in Annotation mode:
class Program
{
static void Main(string[] args)
{
A a = new A();
a.id = 1;
a.b.name = "value";
string json = null;
// json = JsonConvert.SerializeObject(a, new FJson()); without [JsonConverter(typeof(FJson))] annotation workd fine
// json = JsonConvert.SerializeObject(a); StackOverflowException
Console.WriteLine(json);
Console.ReadLine();
}
}
//[JsonConverter(typeof(FJson))] StackOverflowException
public class A
{
public A()
{
this.b = new B();
}
public int id { get; set; }
public string name { get; set; }
public B b { get; set; }
}
public class B
{
public string name { get; set; }
}
public class FJson : JsonConverter
{
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
JToken t = JToken.FromObject(value);
if (t.Type != JTokenType.Object)
{
t.WriteTo(writer);
return;
}
JObject o = (JObject)t;
writer.WriteStartObject();
WriteJson(writer, o);
writer.WriteEndObject();
}
private void WriteJson(JsonWriter writer, JObject value)
{
foreach (var p in value.Properties())
{
if (p.Value is JObject)
WriteJson(writer, (JObject)p.Value);
else
p.WriteTo(writer);
}
}
public override object ReadJson(JsonReader reader, Type objectType,
object existingValue, JsonSerializer serializer)
{
throw new NotImplementedException();
}
public override bool CanConvert(Type objectType)
{
return true; // works for any type
}
}

After reading (and testing) Paul Kiar & p.kaneman solution I'd say it seems to be a challenging task to implement WriteJson. Even though it works for the most cases - there are a few edge cases that are not covered yet.
Examples:
public bool ShouldSerialize*() methods
null values
value types (struct)
json converter attributes
..
Here is (just) another try:
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer) {
if (ReferenceEquals(value, null)) {
writer.WriteNull();
return;
}
var contract = (JsonObjectContract)serializer
.ContractResolver
.ResolveContract(value.GetType());
writer.WriteStartObject();
foreach (var property in contract.Properties) {
if (property.Ignored) continue;
if (!ShouldSerialize(property, value)) continue;
var property_name = property.PropertyName;
var property_value = property.ValueProvider.GetValue(value);
writer.WritePropertyName(property_name);
if (property.Converter != null && property.Converter.CanWrite) {
property.Converter.WriteJson(writer, property_value, serializer);
} else {
serializer.Serialize(writer, property_value);
}
}
writer.WriteEndObject();
}
private static bool ShouldSerialize(JsonProperty property, object instance) {
return property.ShouldSerialize == null
|| property.ShouldSerialize(instance);
}

Json.NET does not have convenient support for converters that call JToken.FromObject to generate a "default" serialization and then modify the resulting JToken for output - precisely because the StackOverflowException due to recursive calls to JsonConverter.WriteJson() that you have observed will occur.
One workaround is to temporarily disable the converter in recursive calls using a thread static Boolean. A thread static is used because, in some situations including asp.net-web-api, instances of JSON converters will be shared between threads. In such situations disabling the converter via an instance property will not be thread-safe.
public class FJson : JsonConverter
{
[ThreadStatic]
static bool disabled;
// Disables the converter in a thread-safe manner.
bool Disabled { get { return disabled; } set { disabled = value; } }
public override bool CanWrite { get { return !Disabled; } }
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
JToken t;
using (new PushValue<bool>(true, () => Disabled, (canWrite) => Disabled = canWrite))
{
t = JToken.FromObject(value, serializer);
}
if (t.Type != JTokenType.Object)
{
t.WriteTo(writer);
return;
}
JObject o = (JObject)t;
writer.WriteStartObject();
WriteJson(writer, o);
writer.WriteEndObject();
}
private void WriteJson(JsonWriter writer, JObject value)
{
foreach (var p in value.Properties())
{
if (p.Value is JObject)
WriteJson(writer, (JObject)p.Value);
else
p.WriteTo(writer);
}
}
public override object ReadJson(JsonReader reader, Type objectType,
object existingValue, JsonSerializer serializer)
{
throw new NotImplementedException();
}
public override bool CanConvert(Type objectType)
{
return true; // works for any type
}
}
public struct PushValue<T> : IDisposable
{
Action<T> setValue;
T oldValue;
public PushValue(T value, Func<T> getValue, Action<T> setValue)
{
if (getValue == null || setValue == null)
throw new ArgumentNullException();
this.setValue = setValue;
this.oldValue = getValue();
setValue(value);
}
#region IDisposable Members
// By using a disposable struct we avoid the overhead of allocating and freeing an instance of a finalizable class.
public void Dispose()
{
if (setValue != null)
setValue(oldValue);
}
#endregion
}
Having done this, you can restore the [JsonConverter(typeof(FJson))] to your class A:
[JsonConverter(typeof(FJson))]
public class A
{
}
Demo fiddle #1 here.
A second, simpler workaround for generating a default JToken representation for a type with a JsonConverter applied takes advantage fact that a converter applied to a member supersedes converters applied to the type, or in settings. From the docs:
The priority of which JsonConverter is used is the JsonConverter defined by attribute on a member, then the JsonConverter defined by an attribute on a class, and finally any converters passed to the JsonSerializer.
Thus it is possible to generate a default serialization for your type by nesting it inside a DTO with a single member whose value is an instance of your type and has a dummy converter applied which does nothing but fall back to to default serialization for both reading and writing.
The following extension method and converter do the job:
public static partial class JsonExtensions
{
public static JToken DefaultFromObject(this JsonSerializer serializer, object value)
{
if (value == null)
return JValue.CreateNull();
var dto = Activator.CreateInstance(typeof(DefaultSerializationDTO<>).MakeGenericType(value.GetType()), value);
var root = JObject.FromObject(dto, serializer);
return root["Value"].RemoveFromLowestPossibleParent() ?? JValue.CreateNull();
}
public static object DefaultToObject(this JToken token, Type type, JsonSerializer serializer = null)
{
var oldParent = token.Parent;
var dtoToken = new JObject(new JProperty("Value", token));
var dtoType = typeof(DefaultSerializationDTO<>).MakeGenericType(type);
var dto = (IHasValue)(serializer ?? JsonSerializer.CreateDefault()).Deserialize(dtoToken.CreateReader(), dtoType);
if (oldParent == null)
token.RemoveFromLowestPossibleParent();
return dto == null ? null : dto.GetValue();
}
public static JToken RemoveFromLowestPossibleParent(this JToken node)
{
if (node == null)
return null;
// If the parent is a JProperty, remove that instead of the token itself.
var contained = node.Parent is JProperty ? node.Parent : node;
contained.Remove();
// Also detach the node from its immediate containing property -- Remove() does not do this even though it seems like it should
if (contained is JProperty)
((JProperty)node.Parent).Value = null;
return node;
}
interface IHasValue
{
object GetValue();
}
[JsonObject(NamingStrategyType = typeof(DefaultNamingStrategy), IsReference = false)]
class DefaultSerializationDTO<T> : IHasValue
{
public DefaultSerializationDTO(T value) { this.Value = value; }
public DefaultSerializationDTO() { }
[JsonConverter(typeof(NoConverter)), JsonProperty(ReferenceLoopHandling = ReferenceLoopHandling.Serialize)]
public T Value { get; set; }
object IHasValue.GetValue() { return Value; }
}
}
public class NoConverter : JsonConverter
{
// NoConverter taken from this answer https://stackoverflow.com/a/39739105/3744182
// To https://stackoverflow.com/questions/39738714/selectively-use-default-json-converter
// By https://stackoverflow.com/users/3744182/dbc
public override bool CanConvert(Type objectType) { throw new NotImplementedException(); /* This converter should only be applied via attributes */ }
public override bool CanRead { get { return false; } }
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer) { throw new NotImplementedException(); }
public override bool CanWrite { get { return false; } }
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer) { throw new NotImplementedException(); }
}
And then use it in FJson.WriteJson() as follows:
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
JToken t = serializer.DefaultFromObject(value);
// Remainder as before
if (t.Type != JTokenType.Object)
{
t.WriteTo(writer);
return;
}
JObject o = (JObject)t;
writer.WriteStartObject();
WriteJson(writer, o);
writer.WriteEndObject();
}
The advantages and disadvantages of this approach are that:
It doesn't rely on recursively disabling the converter, and so works correctly with recursive data models.
It doesn't require re-implementing the entire logic of serializing an object from its properties.
It serializes to and deserializes from an intermediate JToken representation. It is not appropriate for use when attempt to stream a default serialization directly to and from a the incoming JsonReader or JsonWriter.
Demo fiddle #2 here.
Notes
Both converter versions only handle writing; reading is not implemented.
To solve the equivalent problem during deserialization, see e.g. Json.NET custom serialization with JsonConverter - how to get the "default" behavior.
Your converter as written creates JSON with duplicated names:
{
"id": 1,
"name": null,
"name": "value"
}
This, while not strictly illegal, is generally considered to be bad practice and so should probably be avoided.

I didn't like the solution posted above so I worked out how the serializer actually serialized the object and tried to distill it down to the minimum:
public override void WriteJson( JsonWriter writer, object value, JsonSerializer serializer )
{
JsonObjectContract contract = (JsonObjectContract)serializer.ContractResolver.ResolveContract( value.GetType() );
writer.WriteStartObject();
foreach ( var property in contract.Properties )
{
writer.WritePropertyName( property.PropertyName );
writer.WriteValue( property.ValueProvider.GetValue(value));
}
writer.WriteEndObject();
}
No stack overflow problem and no need for a recursive disable flag.

I can't comment yet, so sorry for that...but I just wanted to add something to the solution provided by Paul Kiar. His solution really helped me out.
The code of Paul is short and simply works without any custom building of objects.
The only addition I would like to make is to insert a check if the property is ignored. If it is set to be ignored then skip the write for that property:
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
JsonObjectContract contract = (JsonObjectContract)serializer.ContractResolver.ResolveContract(value.GetType());
writer.WriteStartObject();
foreach (var property in contract.Properties)
{
if (property.Ignored)
continue;
writer.WritePropertyName(property.PropertyName);
writer.WriteValue(property.ValueProvider.GetValue(value));
}
writer.WriteEndObject();
}

By placing the attribute on class A, it is being called recursively. The first line in WriteJson override is again calling the serializer on class A.
JToken t = JToken.FromObject(value);
This causes a recursive call and hence the StackOverflowException.
From your code, I think you are trying to flatten the heirarchy. You can probably achieve this by putting the converter attribute on the property B, which will avoid the recursion.
//remove the converter from here
public class A
{
public A()
{
this.b = new B();
}
public int id { get; set; }
public string name { get; set; }
[JsonConverter(typeof(FJson))]
public B b { get; set; }
}
Warning: The Json you get here will have two keys called "name" one from class A and the other from class B.

Skip json elements deserialization on condition

have the following problem. There is a bunch of quite large json objects which I need to have as objects. But there are conditions, when I actually don't need some of them, for example, if one of inner fields is "isDeleted:true". For now I have to first automatically through JsonConvert.DeserializeObject<..>(stringJson) all the array of data and then choose in filter necessary. What I'd like to achieve is to skip some whole elements when condition isn't true. Are there any ways to achieve this beside manual conversion?
UPDATE:
json looks something like this:
{
results:[
around 20 fields here with inner objects,
isDeleted:"true",
user:{
another dozen of fields
},
images:[{..},{..}]
]}

According to the docs, JsonReader:
Represents a reader that provides fast, non-cached, forward-only access to serialized JSON data.
Thus, in order to skip deleted items when deserializing a collection, you're going to need to read the JSON for each collection entry into memory somehow in order to check the isDeleted property. If your collection item class is so heavy that you don't want to create one when not absolutely required, you could use a custom JsonConverter to load each entry into a JObject container, then check that for isDeleted:
public class DeletetedItemSkippingCollectionConverter<T> : JsonConverter
{
public override bool CanConvert(Type objectType)
{
return objectType == typeof(List<T>);
}
public override bool CanWrite { get { return false; } }
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
var list = (existingValue as List<T>);
if (list == null)
list = new List<T>();
if (reader.TokenType != JsonToken.StartArray)
return list;
while (reader.Read())
{
if (reader.TokenType == JsonToken.EndArray)
break;
var obj = JObject.Load(reader);
var deleted = obj["isDeleted"];
if (deleted != null)
{
if (deleted.Type == JTokenType.Boolean && (bool)deleted)
continue;
}
if (obj != null)
{
list.Add((T)obj.ToObject<T>(serializer));
}
}
return list;
}
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}
Then apply that to any List<T> property you want to filter:
[DataContract]
public class RootObject
{
public RootObject()
{
this.ResultList = new List<Item>();
}
[DataMember(Name="results")]
[JsonConverter(typeof(DeletetedItemSkippingCollectionConverter<Item>))]
public List<Item> ResultList { get; set; }
}
Alternatively, you could load each item in the collection into its "final" representation, check it for being deleted, and dispose of it if so. Assuming your objects all implement an interface IsDeleted, you could do:
public interface IIsDeleted
{
bool IsDeleted { get; } // Corresponds it "isDeleted" property in the JSON.
}
public class IsDeletetedItemSkippingCollectionConverter<T> : JsonConverter where T : IIsDeleted
{
public override bool CanConvert(Type objectType)
{
return objectType == typeof(List<T>);
}
public override bool CanWrite { get { return false; } }
public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
{
var list = (existingValue as List<T>);
if (list == null)
list = new List<T>();
if (reader.TokenType != JsonToken.StartArray)
return list;
while (reader.Read())
{
if (reader.TokenType == JsonToken.EndArray)
break;
var obj = serializer.Deserialize<T>(reader);
if (obj == null || obj.IsDeleted)
{
var disposable = obj as IDisposable;
if (disposable != null)
disposable.Dispose();
continue;
}
list.Add(obj);
}
return list;
}
public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
{
throw new NotImplementedException();
}
}