Develop Reference

Map [name,value] string values to class without reflection

I'm having a huge performance issue about mapping string property names and string property values to classes using reflection.
My issue now:
public class Person
{
public string Property1 { get; set; }
public string Property2 { get; set; }
public string Property3 { get; set; }
public string Property4 { get; set; }
// My class has around 100 properties
public string Property100 { get; set; }
}
I am mapping a key value pair collection to the class using reflection
[{"Property1": "some value"}, {"Property2": "something else"},{"Property3","Property4","value" }.....{"Property100","val"}]
It got to the point that I am now mapping around 10 000 class instances using reflection and the performance is to say it lightly bad.
Any ideas for eliminating the reflection would be greatly appreciated.

I see two options, if you need to avoid reflection for tasks like this(when code could be programatically generated).
First is Expressions I use it often, e.g. I saw some people write something like this
public class A
{
public Prop1 ...
....
public Prop100
public override ToString() => $"{nameof(Prop1)}={Prop1};...";
and so for all 100 properties, and always doing this manually.
And with Expression it can be easily automated, you just need to generate Expression for String.Concat and pass list of properties and names there.
For your example, it is not clear what are your data. How do you do lookup in the list?
Let's assume there is a dictionary<string,string>(you can transform your list of tuples to a dictionary), and all properties are strings as well.
Then we would need to generate a list assignment expressions like this
if(data.ContainsKey("Prop1")) result.Prop1 = data["Prop1"];
And the code would be complicated, anyway it would look like this
private static class CompiledDelegate<T>
{
public static Action<T, Dictionary<string, string>> initObject;
static CompiledDelegate()
{
var i = Expression.Parameter(typeof(Dictionary<string, string>), "i");
var v = Expression.Parameter(typeof(T), "v");
var propertyInfos = typeof(T).GetProperties().ToArray();
var t = new Dictionary<string, string>();
var contains = typeof(Dictionary<string, string>).GetMethod(nameof(Dictionary<string, string>.ContainsKey));
var getter = typeof(Dictionary<string, string>).GetProperties().First(x => x.GetIndexParameters().Length > 0);
var result = new List<Expression>();
foreach (var propertyInfo in propertyInfos)
{
var cst = Expression.Constant(propertyInfo.Name);
var assignExpression =
Expression.IfThen(Expression.Call(i, contains, cst),
Expression.Assign(Expression.PropertyOrField(v, propertyInfo.Name), Expression.MakeIndex(i, getter, new[] { cst })));
result.Add(assignExpression);
}
var block = Expression.Block(result);
initObject = Expression.Lambda<Action<T, Dictionary<string, string>>>(block, new ParameterExpression[] { v, i }).Compile();
}
}
It is an example, it would fail if there were non-string properties.
And it could be used like this
static void Main(string[] args)
{
var tst = new Test();
CompiledDelegate<Test>.initObject(tst, new Dictionary<string, string>
{
{ "S3", "Value3" },
{ "S2", "Value2" },
});
CompiledDelegate<Test>.initObject(tst, new Dictionary<string, string>
{
{ "S3", "Value3" },
{ "S1", "Value1" },
});
Console.ReadKey();
}
The second option is, actually, what it should be ideally imlemented like Using source generators I think such things do have to be done just in build time.
There is a lot of articles on msdn, for instance with samples. But it turned out to be not very easy to implement, even just a sample.
I can say, it didn't work for me, while I tried to do it according to samples.
In order to get it work I had to change TargetFramework to netstandard2.0, do something else...
But after all, when build was green, Visual Studio still showed an error.
Ok, it disappeared after VS restart, but still, that doesn't look very usable.
So, this is a generator, that creates a converter for every class with attribute.
It is again a sample, it doesn't check many things.
[Generator]
public class ConverterGenerator : ISourceGenerator
{
private static string mytemplate = #"using System.Collections.Generic;
using {2};
namespace GeneratedConverters
{{
public static class {0}Converter
{{
public static {0} Convert(Dictionary<string, string> data)
{{
var result = new {0}();
{1}
return result;
}}
}}
}}";
public static string GetNamespaceFrom(SyntaxNode s)
{
if (s.Parent is NamespaceDeclarationSyntax namespaceDeclarationSyntax)
{
return namespaceDeclarationSyntax.Name.ToString();
}
if (s.Parent == null)
return "";
return GetNamespaceFrom(s.Parent);
}
public void Execute(GeneratorExecutionContext context)
{
GetMenuComponents(context, context.Compilation);
}
private static void GetMenuComponents(GeneratorExecutionContext context, Compilation compilation)
{
var allNodes = compilation.SyntaxTrees.SelectMany(s => s.GetRoot().DescendantNodes());
var allClasses = allNodes.Where(d => d.IsKind(SyntaxKind.ClassDeclaration)).OfType<ClassDeclarationSyntax>();
var classes = allClasses
.Where(c => c.AttributeLists.SelectMany(a => a.Attributes).Select(a => a.Name).Any(s => s.ToString().Contains("DictionaryConverter")))
.ToImmutableArray();
foreach (var item in classes.Distinct().Take(1))
{
context.AddSource(item.Identifier.Text + "Converter", String.Format(mytemplate, item.Identifier.Text, SourceText.From(GenerateProperties(item)), GetNamespaceFrom(item)));
}
}
private static string GenerateProperties(ClassDeclarationSyntax s)
{
var properties = s.Members.OfType<PropertyDeclarationSyntax>();
return String.Join(Environment.NewLine,
properties.Select(p =>
{
var name = p.Identifier.Text;
return $"if(data.ContainsKey(\"{name}\")) result.{name} = data[\"{name}\"];";
}));
}
public void Initialize(GeneratorInitializationContext context)
{
}
}
and
static void Main(string[] args)
{
var t1 = GeneratedConverters.TestConverter.Convert(new Dictionary<string, string>
{
{ "S3", "Value3" },
{ "S2", "Value2" },
});
}

Best performance without reflection would be manual mapping.
It seems your key/value pair collection is regular JSON. So you could use the JSONTextReader from JSON.NET and read the string. Then manually map the JSON properties to the class properties.
Like so:
JsonTextReader reader = new JsonTextReader(new StringReader(jsonString));
while (reader.Read())
{
if (reader.Value != null)
{
// check reader.Value.ToString() and assign to correct class property
}
}
More info can be found on the JSON.NET website : https://www.newtonsoft.com/json/help/html/ReadingWritingJSON.htm

IOS 14 request limited photo access

I'm trying to use PHPickerController and access PHAsset to get file name and file size but the PHAsset are null
var config = new PHPickerConfiguration(PHPhotoLibrary.SharedPhotoLibrary) {
Filter = PHPickerFilter.ImagesFilter,
SelectionLimit = 1
};
var picker= new PHPickerViewController(config) {
ModalPresentationStyle = UIModalPresentationStyle.Popover,
Delegate = new ImagePickerDelegate((fileSize, fileName, url) => {
})
};
ViewController.PresentViewController(picker, true, null);
public class ImagePickerDelegate : PHPickerViewControllerDelegate
{
public ImagePickerDelegate(Action<int, string, string> action)
{
Action = action;
}
public Action<int, string, string> Action { get; }
public override void DidFinishPicking(PHPickerViewController picker, PHPickerResult[] results)
{
picker.DismissViewController(true, null);
foreach (var result in results)
{
var asset = PHAsset.FetchAssets(result.AssetIdentifier, null)?.firstObject as PHAsset;
// The asset are null
var fileSize = asset.ValueForKey((NSString)"fileSize");
}
}
}
As you can see in the image the request dialog show and code are not pause on following line
var asset = PHAsset.FetchAssets(result.AssetIdentifier, null)?.firstObject as PHAsset;
and return null

You could use FetchAssetsUsingLocalIdentifiers method to get PHAsset object, then it will return value.
Sample code as follows:
public override void DidFinishPicking(PHPickerViewController picker, PHPickerResult[] results)
{
picker.DismissViewController(true, null);
foreach (var result in results)
{
var refID = result.AssetIdentifier;
string[] refIDs = new string[] { refID };
var asset = PHAsset.FetchAssetsUsingLocalIdentifiers(refIDs, null)?.firstObject as PHAsset;
// var fileSize = asset.ValueForKey((NSString)"fileSize");
}
}
Also could have a look at this native code link.

Merge Sparse Data into Dictionary using Json.NET PopulateObject

I would like to load sparse data in JSON format to get a result with missing data filled in with defaults, but my defaults include predefined instances of an extensible set rather than just fixed fields.
For (arbitrary) example,
Types
class Link
{
public string Addr;
public short Port;
public Link() { Addr = "0.0.0.0"; Port = 80; }
public override string ToString() { return Addr + ":" + Port.ToString(); }
}
class Targets
{
public Link Fixed;
public Dictionary<string, Link> Variable;
public Targets()
{
Fixed = new Link() { Addr = "192.168.0.1" };
Variable = new Dictionary<string, Link>
{
["Common"] = new Link() { Addr = "192.168.0.2" }
};
}
public override string ToString()
{
var result = new System.Text.StringBuilder();
result.Append("Fixed").Append('=').Append(Fixed)
.Append(' ');
foreach (var link in Variable)
{
result.Append(link.Key).Append('=').Append(link.Value)
.Append(' ');
}
return result.ToString();
}
}
Usage
var targets = new Targets();
string json = #"{
'Fixed': { 'Port':12345 },
'Variable': {
'Common': { 'Port':12345 }
}
}";
Newtonsoft.Json.JsonConvert.PopulateObject(json, targets);
Console.WriteLine(targets);
Outputs Fixed=192.168.0.1:12345 Common=0.0.0.0:12345 rather than the desired Fixed=192.168.0.1:12345 Common=192.168.0.2:12345.
This shows that the desired merge logic works for fixed properties, but not for items in a Dictionary despite the fact that the Dictionary will otherwise serialize/deserialize just like a type with fixed properties.

Took me a while to figure this out. Json.NET has a dedicated function for merging two JObjects together. Here's your example modified to use this method:
using Newtonsoft.Json;
using Newtonsoft.Json.Converters;
using Newtonsoft.Json.Linq;
using Newtonsoft.Json.Serialization;
using System;
using System.Collections.Generic;
namespace ConsoleApp3
{
class Link
{
public string Addr;
public short Port;
public Link() { Addr = "0.0.0.0"; Port = 80; }
public override string ToString() { return Addr + ":" + Port.ToString(); }
}
class Targets
{
public Link Fixed;
public Dictionary<string, Link> Variable;
public Targets()
{
Fixed = new Link() { Addr = "192.168.0.1" };
Variable = new Dictionary<string, Link>
{
["Common"] = new Link() { Addr = "192.168.0.2" },
["Common2"] = new Link() { Addr = "192.168.0.25" }
};
}
public override string ToString()
{
var result = new System.Text.StringBuilder();
result.Append("Fixed").Append('=').Append(Fixed)
.Append(' ');
foreach (var link in Variable)
{
if (link.Key != "Variable")
result.Append(link.Key).Append('=').Append(link.Value)
.Append(' ');
}
return result.ToString();
}
}
class Program
{
static void Main(string[] args)
{
var targets = new Targets();
JObject o1 = JObject.Parse( #"{
'Fixed': { 'Port':12345 },
'Variable': {
'Common': { 'Port':12345 }
}
}");
JObject o2 = JObject.FromObject(targets);
o2.Merge(o1, new JsonMergeSettings
{
// union array values together to avoid duplicates
MergeArrayHandling = MergeArrayHandling.Union
});
string json = o2.ToString();
Console.WriteLine(json);
JsonConvert.PopulateObject(json, targets);
Console.WriteLine(targets);
Console.ReadKey();
}
}
}
The output is:
{
"Fixed": {
"Addr": "192.168.0.1",
"Port": 12345
},
"Variable": {
"Common": {
"Addr": "192.168.0.2",
"Port": 12345
},
"Common2": {
"Addr": "192.168.0.25",
"Port": 80
}
}
}
Fixed=192.168.0.1:12345 Common=192.168.0.2:12345 Common2=192.168.0.25:80
EDIT by OP: Refined into extension methods without extra ToString/deserialization:
static class SerializerExtensions
{
public static T MergeObject<T>(this JsonSerializer serializer, JsonReader json, T target)
{
JObject o1 = JObject.FromObject(target, serializer);
JObject o2 = serializer.Deserialize(json) as JObject;
o1.Merge(o2, new JsonMergeSettings
{ // union array values together to avoid duplicates
MergeArrayHandling = MergeArrayHandling.Union,
// an explicit null removes an existing item
MergeNullValueHandling = MergeNullValueHandling.Merge,
});
serializer.Populate(o1.CreateReader(), target);
return target;
}
public static T MergeObject<T>(this JsonSerializer serializer, JsonReader json, JObject template)
{
JObject o1 = template.DeepClone() as JObject;
JObject o2 = serializer.Deserialize(json) as JObject;
o1.Merge(o2, new JsonMergeSettings
{ // union array values together to avoid duplicates
MergeArrayHandling = MergeArrayHandling.Union,
// an explicit null removes an existing item
MergeNullValueHandling = MergeNullValueHandling.Merge,
});
return serializer.Deserialize<T>(o1.CreateReader());
}
}

Using [JsonProperty("name")] in ModelState.Errors

We have a couple of models that override the name via JsonProperty, but this causes an issue when we get validation errors through ModelState. For example:
class MyModel
{
[JsonProperty("id")]
[Required]
public string MyModelId {get;set;}
}
class MyModelController
{
public IHttpActionResult Post([FromBody] MyModel model)
{
if (!ModelState.IsValid)
{
return HttpBadRequest(ModelState);
}
/* etc... */
}
}
The above Post will return the error The MyModelId field is required. which isn't accurate. We'd like this to say The id field is required.. We've attempted using [DataMember(Name="id")] but get the same result.
Question 1: Is there a way we can get ModelState errors to show the JSON property name rather than the C# property name aside from providing our own error messages on every [Required] attribute?
-- Update --
I've been playing around with this and found a "do-it-yourself" method for re-creating the error messages using custom property names. I'm really hoping there's a built-in way to do this, but this seems to do the job...
https://gist.github.com/Blackbaud-JasonTremper/b64dc6ddb460afa1698daa6d075857e4
Question 2: Can ModelState.Key be assumed to match the <parameterName>.<reflectedProperty> syntax or are there cases where this might not be true?
Question 3: Is there an easier way to determine what the JSON parameter name is expected to be rather than searching via reflection on [DataMember] or [JsonProperty] attributes?

Did you try using DisplayName attribute?
displayname attribute vs display attribute
Also, you can assign an error message to [Required] attribute.
[Required(ErrorMessage = "Name is required")]

I also faced this problem, I modified some code from your link to fit my WebAPI. modelState will also store the old key which is the variable name of the model, plus the Json Property names.
First, create the filter ValidateModelStateFilter
Add [ValidateModelStateFilter] above controller method
The filter source code:
public class ValidateModelStateFilter : ActionFilterAttribute
{
public override void OnActionExecuting(HttpActionContext actionContext)
{
var descriptor = actionContext.ActionDescriptor;
var modelState = actionContext.ModelState;
if (descriptor != null)
{
var parameters = descriptor.GetParameters();
var subParameterIssues = modelState.Keys
.Where(s => s.Contains("."))
.Where(s => modelState[s].Errors.Any())
.GroupBy(s => s.Substring(0, s.IndexOf('.')))
.ToDictionary(g => g.Key, g => g.ToArray());
foreach (var parameter in parameters)
{
var argument = actionContext.ActionArguments[parameter.ParameterName];
if (subParameterIssues.ContainsKey(parameter.ParameterName))
{
var subProperties = subParameterIssues[parameter.ParameterName];
foreach (var subProperty in subProperties)
{
var propName = subProperty.Substring(subProperty.IndexOf('.') + 1);
var property = parameter.ParameterType.GetProperty(propName);
var validationAttributes = property.GetCustomAttributes(typeof(ValidationAttribute), true);
var value = property.GetValue(argument);
modelState[subProperty].Errors.Clear();
foreach (var validationAttribute in validationAttributes)
{
var attr = (ValidationAttribute)validationAttribute;
if (!attr.IsValid(value))
{
var parameterName = GetParameterName(property);
// modelState.AddModelError(subProperty, attr.FormatErrorMessage(parameterName));
modelState.AddModelError(parameterName, attr.FormatErrorMessage(parameterName));
}
}
}
}
}
}
}
private string GetParameterName(PropertyInfo property)
{
var dataMemberAttribute = property.GetCustomAttributes<DataMemberAttribute>().FirstOrDefault();
if (dataMemberAttribute?.Name != null)
{
return dataMemberAttribute.Name;
}
var jsonProperty = property.GetCustomAttributes<JsonPropertyAttribute>().FirstOrDefault();
if (jsonProperty?.PropertyName != null)
{
return jsonProperty.PropertyName;
}
return property.Name;
}
}

You can access the parameter type, get the json name, and then replace the property name with the json name. Something like this:
var invalidParameters = (from m in actionContext.ModelState
where m.Value.Errors.Count > 0
select new InvalidParameter
{
ParameterName = m.Key,
ConstraintViolations = (from msg in m.Value.Errors select msg.ErrorMessage).ToArray()
}).AsEnumerable().ToArray();
if (actionContext.ActionDescriptor.Parameters.Count == 1)
{
var nameMapper = new Dictionary<string, string>();
foreach (var property in actionContext.ActionDescriptor.Parameters[0].ParameterType.GetProperties())
{
object[] attributes = property.GetCustomAttributes(typeof(JsonPropertyNameAttribute), false);
if (attributes.Length != 1) continue;
nameMapper.Add(property.Name, ((JsonPropertyNameAttribute) attributes[0]).Name);
}
var modifiedInvalidParameters = new List<InvalidParameter>();
foreach (var invalidParameter in invalidParameters)
{
if(invalidParameter.ParameterName != null && nameMapper.TryGetValue(invalidParameter.ParameterName, out var mappedName))
{
var modifiedConstraintViolations = new List<string>();
foreach (var constraintViolation in invalidParameter.ConstraintViolations ?? Enumerable.Empty<string>())
{
modifiedConstraintViolations.Add(constraintViolation.Replace(invalidParameter.ParameterName, mappedName));
}
modifiedInvalidParameters.Add(new InvalidParameter
{
ParameterName = mappedName,
ConstraintViolations = modifiedConstraintViolations.ToArray()
});
}
else
{
modifiedInvalidParameters.Add(invalidParameter);
}
}
invalidParameters = modifiedInvalidParameters.ToArray();
}
public struct InvalidParameter
{
[JsonPropertyName("parameter_name")]
public string? ParameterName { get; set; }
[JsonPropertyName("constraint_violations")]
public string[]? ConstraintViolations { get; set; }
}

JSON with dynamic Root-Object

problably I'm not experienced enought and my question is kind of dumb:
For learning purposes I'm trying to connect to a REST-Service, which delivers JSON-Data.
From what I've learned, the purpose of JSON is to deliver the same data to any possible client without having a State of itself.
My code is looking like this:
public static void DoSomething()
{
HttpClient client = new HttpClient();
client.BaseAddress = new Uri("SomeUrl"));
// Add an Accept header for JSON format.
client.DefaultRequestHeaders.Accept.Add(new MediaTypeWithQualityHeaderValue("application/json"));
// List data response.
HttpResponseMessage response = client.GetAsync("").Result;
if (response.IsSuccessStatusCode)
{
Task<Stream> readTask = response.Content.ReadAsStreamAsync();
readTask.ContinueWith(task =>
{
DataContractJsonSerializer ser = new DataContractJsonSerializer(typeof(RootObject));
using (Stream result = task.Result)
{
result.Position = 0;
RootObject obj = (RootObject)ser.ReadObject(result);
}
});
}
else
{
Console.WriteLine("{0} ({1})", (int)response.StatusCode, response.ReasonPhrase);
}
}
public class Sum
{
public int id { get; set; }
public string name { get; set; }
public int profileIconId { get; set; }
public int summonerLevel { get; set; }
public long revisionDate { get; set; }
}
public class RootObject
{
public Sum khalgor { get; set; }
}
But here's my Problem: I've created this classes "Sum" and "RootObject" by using the Website http://json2csharp.com/, the JSON-String is looking like this:
{"khalgor":{"id":23801741,"name":"Khalgor","profileIconId":7,"summonerLevel":30,"revisionDate":1396876104000}}
The Problem: The Name "Khalgor" seems to be used as a Root-Object, but it's a Name. So if I'd like to user for another Name, I'd have to user another RootObject.
It does not make that much sense to create such a Structure, so my question: What's the best practice here? Do I map this RootObject/Property to another object manually? Do I use some Reflection to dynamically create an Property or rename it?
As usual, thanks a lot for all Responses
Matthias
Edit:
I tinkered arround a bit and that's my first idea of a solution:
public static class LOLObjectFactory
{
public static ILOLObject Create(string jsonString)
{
JavaScriptSerializer jss = new JavaScriptSerializer();
Dictionary<String, object> entry = (jss.Deserialize<dynamic>(jsonString) as Dictionary<string, object>).First().Value as Dictionary<String, object>;
Type selectedType = null;
List<string> fieldNames = entry.Select(f => f.Key).OrderBy(f => f).ToList();
Type[] types = typeof(ILOLObject).Assembly.GetTypes();
foreach(var type in types)
{
List<string> typeProperties = type.GetProperties().Select(f => f.Name).OrderBy(f => f).ToList();
if (fieldNames.SequenceEqual(typeProperties) && typeof(ILOLObject).IsAssignableFrom(type))
{
selectedType = type;
break;
}
}
ILOLObject result = System.Activator.CreateInstance(selectedType) as ILOLObject;
foreach(var prop in result.GetType().GetProperties())
{
prop.SetValue(result, entry.First(f => f.Key == prop.Name).Value);
}
return result;
}
}
So all the objects I have have the ILOLObject implemented. I'm sure it's not working for everything, but I guess that would be a good approach?
Edit2: Just by looking at it I see I'll have a lot of work to do, but I think the idea behind it is quite clear.

I think your best bet for json "fragments" is to deserialize into a dynamic object:
dynamic stuff = JsonConvert.DeserializeObject(inputData);
Then you can deserialize parts that make sense into proper .NET objects.
SomeObject o = JsonConvert.DeserializeObject<SomeObject>(stuff["someProperty"].ToString());

If you want to ignore the root altogether (e.g. it changes its name everytime) use Json.NET to parse it into an object and ignore the topmost element. Example:
JObject obj = JObject.Parse(json);
if (obj != null)
{
var root = obj.First;
if (root != null)
{
var sumJson = root.First;
if (sumJson != null)
{
var sum = sumJson.ToObject<Sum>();
}
}
}