We Keep Coding

Deep copying a Func within an object in C#

I am building a list of unit tests, which are organised as a list of objects, each of which contain the test method to be executed as a Func. Each object has a variable which is within scope of the Func and is used by it. The variable is not passed in as a parameter.
Iterating over the list and running all the tests runs fine, but is is possible to copy a Func from one object, -breaking the reference to that object-, and assign it to a new object? I assume this is possible somehow by creating a Deep Copy, but my attempt using BinaryFormatter has not worked, any tips would be appreciated!
I have a simplified forms application as follows to illustrate my problem:
using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Data;
using System.Drawing; using
using System.IO;
using System.Linq;
using System.Runtime.Serialization.Formatters.Binary;
using System.Text;
using System.Threading.Tasks; using System.Windows.Forms;
namespace WindowsFormsApplication4 {
public partial class Form1 : Form
{
public Form1()
{
InitializeComponent();
}
public static object DeepClone(object obj)
{
object objResult = null;
using (MemoryStream ms = new MemoryStream())
{
BinaryFormatter bf = new BinaryFormatter();
bf.Serialize(ms, obj);
ms.Position = 0;
objResult = bf.Deserialize(ms);
}
return objResult;
}
[Serializable]
public class POCOwithFunc {
public POCOwithFunc(Func<string> myfunc)
{
mqi = myfunc;
}
public POCOwithFunc() { }
public Func<string> mqi;
public object parm;
}
private void button1_Click(object sender, EventArgs e)
{
List<POCOwithFunc> testList = new List<POCOwithFunc>();
for (int x = 0; x < 5; x++)
{
var pc = new POCOwithFunc();
pc.parm = x;
pc.mqi = delegate()
{
var rrq = pc.parm;
return "result: " + pc.parm;
};
testList.Add(pc);
}
String output = "";
foreach (var test in testList)
{
output += test.mqi() + "\r\n";
}
//output:
//result: 0
//result: 1
//result: 2
//result: 3
//result: 4
var pocoToBeCopied = testList[2];
var newpoco = new POCOwithFunc();
newpoco.parm = 10;
newpoco.mqi = pocoToBeCopied.mqi;
var res = newpoco.mqi(); //returns 2
newpoco = (POCOwithFunc)DeepClone(pocoToBeCopied); //fails
}
} }

This is the first time I'm hearing about deep copy a delegate (which would not work, as the delegate (Func is type of delegate) contain reference to its closure (its environment, which contain any variable that that delegate is using).
I would suggest to change the parameter itself, or, to send it as a parameter (there is a delegate type for it too: Func<object, string >).
(and, I think you should think about redesign the whole thing :-/)

7 years later...
So I had encountered pretty much the same problem, also when writing tests (basically, I had a delegate that captured a lot of info from different parts of the program, and at the end of the program displayed it in a elaborate, error-prone way. I wanted to check that in all cases where captured variable turns to null/other weird values, the delegate handles it gracefully). I had a general idea that "no you can't" isn't completely true, so here goes.
TL;DR:
Closures: compiler generates a class that contains references to all
"captured" variables. This class can be accessed and edited via
reflection, thus "redirecting" the closure to point to another
variable.
Delegate target can also be edited via reflection.
To "deep copy" a delegate within an object you need to change
delegates' Target and (if needed) redirect any fields that point to
original object.
Long version:
Turned out, there is two questions here: first (and the one that was more important for me) - redirecting closures. When compiler detects a delegate that forms a closure that is moved beyond the current scope, a class is generated. This class contains references to the objects that were "captured", so that they don't get garbage collected despite being out of scope. (Also, turned out, the fields of this class are helpfully called by the name of variable that got captured.) You can see the generated class with Reflection:
<!-- language: csharp -->
public static void PrintClass(System.Delegate f)
{
var compilerGeneratedType = f.Target.GetType();
var fields = compilerGeneratedType.GetFields(BindingFlags.Instance | BindingFlags.Public | BindingFlags.NonPublic);
Console.WriteLine(compilerGeneratedType.Name + " has fields:");
foreach (var field in fields)
{
Console.WriteLine(field.FieldType + "\t" + field.Name);
}
}
and the output when used on a classic closure example:
Action a = () => { };
for (int index = 0; index != 3; index++)
{
a += () => { Console.Write(index + " "); };
}
a(); //output: 3 3 3
PrintClass(a);
output of PrintClass function:
<>c__DisplayClass0_1 has fields: System.Int32 index
Modifying that closure can be done with reflection, like that:
static System.Delegate RedirectClosure(System.Delegate f, object newTarget, string originalVariableName)
{
System.Delegate result = (System.Delegate)f.Clone();
var compilerGeneratedType = f.Target.GetType();
var fields = compilerGeneratedType.GetFields();
foreach (var field in fields)
{
if (field.FieldType == newTarget.GetType() && field.Name == originalVariableName)
field.SetValue(result.Target, newTarget);
}
return result;
}
Using that earlier example:
Action a = () => { };
for (int index = 0; index != 3; index++)
{
a += () => { Console.Write(index + " "); };
}
a(); //output: 3 3 3
int j = 42;
RedirectClosure(a, (object)j, "index");
a(); //output: 42 42 42
Now for this specific example (delegate is inside a class, and it captured a field of this same class), second problem: the delegate's Target also needs to be changed. For this I modified the copying function to also redirect delegate's target. (Using .Clone() on delegate or straight-up copying it has the side effect: if I copy delegate from source to destination and modify only its "captured object" field to point to second destination - source delegate also starts pointing there. I guess they still share a reference to same Target.) Modified function (copies the delegate and redirects to new object):
static System.Delegate CopyDelegateAndRedirectClosure<T1>(System.Delegate f, T1 originalTarget, T1 newTarget)
{
System.Delegate result = (System.Delegate)f.Clone();
// I bet there is a better way to get a copy then this =(
var serialized = JsonSerializer.Serialize(result.Target);
var deserialized = JsonSerializer.Deserialize(serialized, result.Target.GetType());
var targetField = result.GetType().GetField("_target", BindingFlags.Instance | BindingFlags.NonPublic);
targetField.SetValue(result,deserialized);
var compilerGeneratedType = f.Target.GetType();
var fields = compilerGeneratedType.GetFields();
foreach (var field in fields)
{
if (field.FieldType == originalTarget.GetType() && field.GetValue(f.Target) == (object)originalTarget)
field.SetValue(result.Target, newTarget);
}
return result;
}
And here is an example of using it on the same situation the original poster had.
Test class:
class A
{
public int param;
public Func<int> del;
}
"Deep copying" the delegate from one instance to another:
var destination = new A { param = 10 };
var source = new A { param = 2 };
source.del = () => { return source.param + 100; };
Console.WriteLine($"source: param = {source.param}, del() result = {source.del()}");
// output:
// source: param = 2, del() result = 102
destination.del = (System.Func<int>)CopyDelegateAndRedirectClosure(source.del, source, destination);
Console.WriteLine($"destination: param = {destination.param}, del() result = {destination.del()}");
Console.WriteLine($"source: param = {source.param}, del() result = {source.del()}");
// output:
// destination: destination: param = 10, del() result = 110
// source: param = 2, del() result = 102
So there it is: copied the delegate from one instance to other, and it now operates on the new instance. The old one is unaffected.
Now again to "why would one do that" - yep, better architecture would have prevented me from ever finding this question. However, this did allow me to write a test replacing all captured variables of a specific type with "broken" value, and "bad architecture + tests" is better then "bad architecture + no tests".

Defining a method with for loops and conditional statements

I have to define a method with Reflection.Emit that is rather complex, because I have to do a for loop on a field and have a condition with break and return. My method that I want to recreate with reflection looks like this in regular code:
override int GetKeyImpl(Type obj0)
{
int answer = -1;
for(int i = 0; i < knownTypes.length; i++){
if(knowntypes[i] == obj0){
answer = i;
break;
}
}
return answer;
}
My idea to solve this problem was to generate a method with reflection that redirects the call to my original method and returns the int.
I need to know how to do a for loop and breaks with OpCodes to recreate the method while doing conditional checks on an array that is inside a class. I've searched for tutorials but didn't find any that go further than addition of two ints.
Edit: Forgot to mention it, I'm using IKVM.Reflection and knownTypes is an array of Type[]. The method that im writing is one that will override an abstract one.

This should reproduce the method you specified:
TypeBuilder type = /* ... */;
FieldInfo knownFields = /* ... */;
// Finding dependencies via reflection
var baseMethod = type.BaseType.GetMethod(
"GetKeyImpl",
BindingFlags.Public | BindingFlags.NonPublic | BindingFlags.Instance);
var typeEqualsOperator = typeof(Type).GetMethod(
"op_Equality",
BindingFlags.Static | BindingFlags.Public | BindingFlags.NonPublic,
null,
new[] { typeof(Type), typeof(Type) },
null);
// Declaring the method
var getKeyImpl = type.DefineMethod(
baseMethod.Name,
baseMethod.Attributes & ~(MethodAttributes.Abstract |
MethodAttributes.NewSlot));
// Setting return type
getKeyImpl.SetReturnType(typeof(int));
// Adding parameters
getKeyImpl.SetParameters(typeof(Type));
getKeyImpl.DefineParameter(1, ParameterAttributes.None, "obj0");
// Override the base method
type.DefineMethodOverride(getKeyImpl, baseMethod);
var il = getKeyImpl.GetILGenerator();
// Preparing locals
var answer = il.DeclareLocal(typeof(int));
var i = il.DeclareLocal(typeof(int));
// Preparing labels
var loopCondition = il.DefineLabel();
var loopIterator = il.DefineLabel();
var returnLabel = il.DefineLabel();
var loopBody = il.DefineLabel();
// Writing body
// answer = -1
il.Emit(OpCodes.Ldc_I4_M1);
il.Emit(OpCodes.Stloc, answer);
// i = 0
il.Emit(OpCodes.Ldc_I4_0);
il.Emit(OpCodes.Stloc, i);
// jump to loop condition
il.Emit(OpCodes.Br_S, loopCondition);
// begin loop body
il.MarkLabel(loopBody);
// if (obj0 != knownTypes[i]) continue
il.Emit(OpCodes.Ldarg_0); // omit if 'knownTypes' is static
il.Emit(OpCodes.Ldfld, knownTypes); // use 'Ldsfld' if 'knownTypes' is static
il.Emit(OpCodes.Ldloc, i);
il.Emit(OpCodes.Ldelem_Ref);
il.Emit(OpCodes.Ldarg_1); // use 'Ldarg_0' if 'knownTypes' is static
il.Emit(OpCodes.Call, typeEqualsOperator);
il.Emit(OpCodes.Brfalse_S, loopIterator);
// answer = i; jump to return
il.Emit(OpCodes.Ldloc, i);
il.Emit(OpCodes.Stloc, answer);
il.Emit(OpCodes.Br_S, returnLabel);
// begin loop iterator
il.MarkLabel(loopIterator);
// i = i + 1
il.Emit(OpCodes.Ldloc, i);
il.Emit(OpCodes.Ldc_I4_1);
il.Emit(OpCodes.Add);
il.Emit(OpCodes.Stloc, i);
// begin loop condition
il.MarkLabel(loopCondition);
// if (i < knownTypes.Length) jump to loop body
il.Emit(OpCodes.Ldloc, i);
il.Emit(OpCodes.Ldarg_0); // omit if 'knownTypes' is static
il.Emit(OpCodes.Ldfld, knownTypes); // use 'Ldsfld' if 'knownTypes' is static
il.Emit(OpCodes.Ldlen);
il.Emit(OpCodes.Conv_I4);
il.Emit(OpCodes.Blt_S, loopBody);
// return answer
il.MarkLabel(returnLabel);
il.Emit(OpCodes.Ldloc, answer);
il.Emit(OpCodes.Ret);
// Finished!
The decompiled results are as expected:
override int GetKeyImpl(Type obj0)
{
for (int i = 0; i < this.knownTypes.Length; i++)
{
if (this.knownTypes[i] == obj0)
return i;
}
return -1;
}
If you have access to .NET Reflector, there is a Reflection.Emit Language Add-In that may interest you. Alternatively, write a prototype in C# code, and then run it through a disassembler to see the raw IL.
If it had been okay to make the method static (and accept knownTypes as a parameter or make it a static field), then you could have composed the method body using LINQ expression trees. Unfortunately, you cannot compose instance method bodies using this technique; they have to be static. Example:
var method = typeBuilder.DefineMethod(
"GetKeyImpl",
MethodAttributes.Private |
MethodAttributes.Static |
MethodAttributes.HideBySig);
var type = E.Parameter(typeof(Type), "type");
var knownTypes = E.Parameter(typeof(Type[]), "knownTypes");
var answer = E.Variable(typeof(int), "answer");
var i = E.Variable(typeof(int), "i");
var breakTarget = E.Label("breakTarget");
var continueTarget = E.Label("continueTarget");
var returnTarget = E.Label(typeof(int), "returnTarget");
var forLoop = E.Block(
new[] { i },
E.Assign(i, E.Constant(0)),
E.Loop(
E.Block(
E.IfThen(
E.GreaterThanOrEqual(i, E.ArrayLength(knownTypes)),
E.Break(breakTarget)),
E.IfThen(
E.Equal(E.ArrayIndex(knownTypes, i), type),
E.Return(returnTarget, i)),
E.Label(continueTarget),
E.PreIncrementAssign(i))),
E.Label(breakTarget));
var body = E.Lambda<Func<Type, Type[], int>>(
E.Block(
new[] { answer },
E.Assign(answer, E.Constant(-1)),
forLoop,
E.Label(returnTarget, answer)),
type,
knownTypes);
body.CompileToMethod(method);
return method;
The example above accepts knownTypes as the second parameter. Refactoring to read from a static field instead would be straightforward. The decompiled results, again, are as expected:
private static int GetKeyImpl(Type type, Type[] knownTypes)
{
for (int i = 0; i < knownTypes.Length; i++)
{
if (knownTypes[i] == type)
return i;
}
return -1;
}

The easist way to work out how to generate the IL for a method is to create a simple console application that has your method in. Then build it and run ILDasm against it to look at the IL instructions necessary to make up the method.
Once you can see the instructions it shouldn't be too difficult to write code to emit the OpCodes necessary.

Checking for Nulls on DB Record Mapping

How can I check for db null values in the attached code? Please understand I am a new C# convert...
What this code does is takes a IDataReader object and converts and maps it to a strongly-typed list of objects. But what I am finding is it completely errors out when there are null columns returned in the reader.
Converter
internal class Converter<T> where T : new()
{
// Declare our _converter delegate
readonly Func<IDataReader, T> _converter;
// Declare our internal dataReader
readonly IDataReader dataReader;
// Build our mapping based on the properties in the class/type we've passed in to the class
private Func<IDataReader, T> GetMapFunc()
{
// declare our field count
int _fc = dataReader.FieldCount;
// declare our expression list
List<Expression> exps = new List<Expression>();
// build our parameters for the expression tree
ParameterExpression paramExp = Expression.Parameter(typeof(IDataRecord), "o7thDR");
ParameterExpression targetExp = Expression.Variable(typeof(T));
// Add our expression tree assignment to the exp list
exps.Add(Expression.Assign(targetExp, Expression.New(targetExp.Type)));
//does int based lookup
PropertyInfo indexerInfo = typeof(IDataRecord).GetProperty("Item", new[] { typeof(int) });
// grab a collection of column names from our data reader
var columnNames = Enumerable.Range(0, _fc).Select(i => new { i, name = dataReader.GetName(i) }).AsParallel();
// loop through all our columns and map them properly
foreach (var column in columnNames)
{
// grab our column property
PropertyInfo property = targetExp.Type.GetProperty(column.name, BindingFlags.Public | BindingFlags.Instance | BindingFlags.IgnoreCase);
// check if it's null or not
if (property != null)
{
// build our expression tree to map the column to the T
ConstantExpression columnNameExp = Expression.Constant(column.i);
IndexExpression propertyExp = Expression.MakeIndex(paramExp, indexerInfo, new[] { columnNameExp });
UnaryExpression convertExp = Expression.Convert(propertyExp, property.PropertyType);
BinaryExpression bindExp = Expression.Assign(Expression.Property(targetExp, property), convertExp);
// add it to our expression list
exps.Add(bindExp);
}
}
// add the originating map to our expression list
exps.Add(targetExp);
// return a compiled cached map
return Expression.Lambda<Func<IDataReader, T>>(Expression.Block(new[] { targetExp }, exps), paramExp).Compile();
}
// initialize
internal Converter(IDataReader dataReader)
{
// initialize the internal datareader
this.dataReader = dataReader;
// build our map
_converter = GetMapFunc();
}
// create and map each column to it's respective object
internal T CreateItemFromRow()
{
return _converter(dataReader);
}
}
Mapper
private static IList<T> Map<T>(DbDataReader dr) where T : new()
{
try
{
// initialize our returnable list
List<T> list = new List<T>();
// fire up the lamda mapping
var converter = new Converter<T>(dr);
while (dr.Read())
{
// read in each row, and properly map it to our T object
var obj = converter.CreateItemFromRow();
// add it to our list
list.Add(obj);
}
// reutrn it
return list;
}
catch (Exception ex)
{
// make sure this method returns a default List
return default(List<T>);
}
}
I just don't quite understand where the column to typed object happens in here, so I'd try to do it myself... but I just don;t know where it is.
I know this probably won't help much, but the error I am getting is:
Unable to cast object of type 'System.DBNull' to type 'System.String'.
and it happens on the
internal T CreateItemFromRow()
{
return _converter(dataReader); //<-- Here
}
Note
This does not happen if I wrap the columns in the query itself with an ISNULL(column, ''), but I am sure you can understand that this is surely not a solution

The problem lies in the line convertExp = Expression.Convert(propertyExp, property.PropertyType). You can't expect to convert DbNull value to its equivalent in framework type. This is especially nasty when your type is a value type. One option is to check if the read value from db is DbNull.Value and in case yes, you need to find a compatible value yourself. In some cases people are ok with default values of those types in C#. If you have to do this
property = value == DBNull.Value ? default(T): value;
a generic implementation would look like (as far as the foreach in your converter class goes):
foreach (var column in columns)
{
var property = targetExp.Type.GetProperty(
column.name,
BindingFlags.Public | BindingFlags.Instance | BindingFlags.IgnoreCase);
if (property == null)
continue;
var columnIndexExp = Expression.Constant(column.i);
var propertyExp = Expression.MakeIndex(
paramExp, indexerInfo, new[] { columnIndexExp });
var convertExp = Expression.Condition(
Expression.Equal(
propertyExp,
Expression.Constant(DBNull.Value)),
Expression.Default(property.PropertyType),
Expression.Convert(propertyExp, property.PropertyType));
var bindExp = Expression.Assign(
Expression.Property(targetExp, property), convertExp);
exps.Add(bindExp);
}
Now this does an equivalent of
property = reader[index] == DBNull.Value ? default(T): reader[index];
You could avoid the double lookup of the reader by assigning it to a variable and using its value in the conditional check. So this should be marginally better, but a lil' more complex:
foreach (var column in columns)
{
var property = targetExp.Type.GetProperty(
column.name,
BindingFlags.Public | BindingFlags.Instance | BindingFlags.IgnoreCase);
if (property == null)
continue;
var columnIndexExp = Expression.Constant(column.i);
var cellExp = Expression.MakeIndex(
paramExp, indexerInfo, new[] { columnIndexExp });
var cellValueExp = Expression.Variable(typeof(object), "o7thPropValue");
var convertExp = Expression.Condition(
Expression.Equal(
cellValueExp,
Expression.Constant(DBNull.Value)),
Expression.Default(property.PropertyType),
Expression.Convert(cellValueExp, property.PropertyType));
var cellValueReadExp = Expression.Block(new[] { cellValueExp },
Expression.Assign(cellValueExp, cellExp), convertExp);
var bindExp = Expression.Assign(
Expression.Property(targetExp, property), cellValueReadExp);
exps.Add(bindExp);
}
This does the conditional check this way:
value = reader[index];
property = value == DBNull.Value ? default(T): value;

This is one of the most annoying problems in dealing with datasets in general.
The way I normally get around it is to convert the DBNull value to something more useful, like an actual null or even a blank string in some cases. This can be done in a number of ways, but just recently I've taken to using extension methods.
public static T? GetValueOrNull<T>(this object value) where T : struct
{
return value == null || value == DBNull.Value ? (T?) null : (T) Convert.ChangeType(value, typeof (T));
}
A handy extension method for nullable types, so for example:
int? myInt = DataSet.Tables[0].Rows[0]["DBNullInt"].GetValueOrNull<int>();
Or a more generic one to just convert a DBNull in to a null:
public static object GetValueOrNull(this object value)
{
return value == DBNull.Value ? null : value;
}
string myString DataSet.Tables[0].Rows[0]["DBNullString"].GetValueOrNull();
You'll then get a null string, rather than trying to put a DBNull in to a string.
Hopefully that may help you a little.

As I come across this problem recently
both
Expression.TypeIs(propertyExp,typeof(DBNull));
and
Expression.Equal(propertyExp,Expression.Constant(DBNull.Value));
didn't work for me as they did increase memory allocation (which is my primary concern in this case)
here is the benchmark for both mapper approach compare to Dapper on 10K rows query.
TypeIs
and
Equal
so to fix this problem it came out that an IDataRecord is able to call "IsDBNull" to check whether the column in current reader is DBNull or not
and can be write as expression like
var isReaderDbNull = Expression.Call(paramExp, "IsDBNull", null, readerIndex);
finally, I end up with this solution
and now the performance is acceptable again.

Recursive call to generic function with DataRow parameter - MakeGenericMethod Fails

the question is : if i have MyConvertDataRowToEntity(DataRow row )
and I call in with T object from type Parent and inside I call the same function with desendant type Child how should I pass the DataRow parameter ?
The problem is created when Invoke of MakeGenericMethod called.
Did change the type to DataSet , string and String types .
No luck.
(I recognize the children object bu prefix in column names - PrefixDataColumn )
public static T MyConvertDataRowToEntity<T>(DataRow row ) where T : class, new()
{
Type objType = typeof(T);
Type parentObjType = typeof(T);
T obj = Activator.CreateInstance<T>(); //hence the new() contsraint
PropertyInfo propertyGenericType = null;
object childInstance = null;
PropertyInfo property;
string childColumnName = string.Empty ;
foreach (DataColumn column in row.Table.Columns)
{
column.ColumnName = column.ColumnName.Replace("_", "");
string PrefixDataColumn;
if (column.ColumnName.IndexOf(".") > (-1))
{
///gets the prefix that is the same as child entity name
PrefixDataColumn = column.ColumnName.Substring(0, column.ColumnName.IndexOf("."));
///the column name in the child
int length = column.ColumnName.Length - 1;
int start = column.ColumnName.IndexOf(".") + 1;
childColumnName = column.ColumnName.Substring(column.ColumnName.IndexOf(".") + 1);
propertyGenericType = objType.GetProperty(PrefixDataColumn,
BindingFlags.Public | BindingFlags.Instance | BindingFlags.IgnoreCase | BindingFlags.FlattenHierarchy);
parentObjType = objType;
if (propertyGenericType != null)
{
Type childType = propertyGenericType.PropertyType;
objType = childType;
childInstance = Activator.CreateInstance(propertyGenericType.PropertyType);
// get the get method for the property
MethodInfo method = propertyGenericType.GetGetMethod(true);
// get the generic get-method generator
MethodInfo genericHelper = typeof(DataUtil).GetMethod("MyConvertDataRowToEntity", BindingFlags.Public | BindingFlags.Static);
List<Type> signature = new List<Type>();
// first type parameter is type of target object
signature.Add(childType);
//next parameters are real types of method arguments
foreach (ParameterInfo pi in genericHelper.GetParameters())
{
signature.Add(pi.ParameterType);
}
// last parameters are known types of method arguments
signature.AddRange(typeof(T).GetGenericArguments());
// reflection call to the generic get-method generator to generate the type arguments
//MethodInfo constructedHelper = genericHelper.MakeGenericMethod(signature.ToArray());
// reflection call to the generic get-method generator to generate the type arguments
MethodInfo constructedHelper = genericHelper.MakeGenericMethod(childType );
// now call it. The null argument is because it's a static method.
object ret = constructedHelper.Invoke(null, new object[] { method });
// object myObj = method.Invoke(null, row);
//// property.SetValue(obj, MyConvertDataRowToEntity<childInstance>(DataRow row),null);
// childInstance = DataUtil.GetMethod("MyConvertDataRowToEntity").MakeGenericMethod(childType); //MyConvertDataRowToEntity<object>(row);
//childType initializedChild = ;
//property.SetValue(obj, value, null);
//objType = parentObjType;
}
else
{
continue;
}
}
}
return obj;
}
Getting this error :
Object of type 'System.Reflection.RuntimeMethodInfo' cannot be converted to type 'System.Data.DataRow'.
Is there any solution for this ?
p.s.
Narrowed down the code as much as i could.
How can I invoke the Method recursivly with desedant types and pass datarow ?

The reason for the error is because you are calling MyConvertDataRowToEntity<ChildType> but then passing in the getaccessor methodinfo for the property as the parameter instead of a data row containing only those fields.
If you want to continue with the code processing logic you are currently using you would need to construct a new datarow containing the fields you wanted (with the prefix and ".") removed from the start of the column names.
Alternatively you could create a helper method the accepted a column name, the source object and it simply updated the value.
static void UpdateItemProperty<T>(T item, string columnName, object rowValue) {
var prefixColumn=columnName.IndexOf(".")==-1 ? columnName : columnName.Split(".")[0];
var pi = typeof(T).GetProperty(prefixColumn // Add your binding flags);
// if pi==null then there is an error...
if (column.ColumnName.IndexOf(".") == (-1)) { // No Nesting
pi.SetValue(item,rowValue);
return;
}
// Nesting
var child=pi.GetValue(item);
if (child==null) {
// Logic here to get childs type and create an instance then call pi.SetValue with child
}
var remainder=string.Join(',',columnName.Split(".").Skip(1).ToArray());
// make your generic method info for UpdateItemProperty with pi.PropertyType into mi
mi.Invoke(null,new object[] { child,remainder,value };
}

C# using reflection to create a struct

I am currently writing some code to save general objects to XML using reflection in c#.
The problem is when reading the XML back in some of the objects are structs and I can't work out how to initialise the struct. For a class I can use
ConstructorInfo constructor = SomeClass.GetConstructor(Type.EmptyTypes);
however, for a struct, there is no constructor which takes no parameters so the above code sets constructor to null. I also tried
SomeStruct.TypeInitializer.Invoke(null)
but this throws a memberaccessexception. Google gives no promising hits. Any help would be appreciated.

If the values are structs, they're likely to be immutable - so you don't want to call a parameterless constructor, but the one which takes the appropriate values as constructor arguments.
If the structs aren't immutable, then run away from them as fast as possible, if you can... but if you absolutely have to do this, then use Activator.CreateInstance(SomeClass). You'll have to be very careful when you use reflection to set properties or fields on the value type though - without that care, you'll end up creating a copy, changing the value on that copy, and then throwing it away. I suspect that if you work with a boxed version throughout, you'll be okay:
using System;
// Mutable structs - just say no...
public struct Foo
{
public string Text { get; set; }
}
public class Test
{
static void Main()
{
Type type = typeof(Foo);
object value = Activator.CreateInstance(type);
var property = type.GetProperty("Text");
property.SetValue(value, "hello", null);
Foo foo = (Foo) value;
Console.WriteLine(foo.Text);
}
}

CreateInstance will not help you with structs with no explicitly defined constructors.
FormatterServices.GetUninitializedObject(Type type);
This does the trick with blank structs.

Just to add - with immutable structs, you are likely to have to do parameter matching to the constructor. Unfortunately this is tricky when there can be multiple constructs, and especially since some types have a separate static "Create" method instead of a public constructor. But assuming you've done the matching, you can still use Activator.CreateInstance:
Type type = typeof(Padding); // just an example
object[] args = new object[] {1,2,3,4};
object obj = Activator.CreateInstance(type, args);
However - the code to pick a constructor (the above has 3...) isn't easy. You could say "pick the most complex" and then attempt to match parameter names to property names (case insensitive)...
A naïve example:
static void Main() {
Dictionary<string, object> propertyBag =
new Dictionary<string, object>();
// these are the values from your xml
propertyBag["Left"] = 1;
propertyBag["Top"] = 2;
propertyBag["Right"] = 3;
propertyBag["Bottom"] = 4;
// the type to create
Type type = typeof(Padding);
object obj = CreateObject(type, propertyBag);
}
static object CreateObject(Type type, IDictionary<string,object> propertyBag)
{
ConstructorInfo[] ctors = type.GetConstructors();
// clone the property bag and make it case insensitive
propertyBag = new Dictionary<string, object>(
propertyBag, StringComparer.OrdinalIgnoreCase);
ConstructorInfo bestCtor = null;
ParameterInfo[] bestParams = null;
for (int i = 0; i < ctors.Length; i++)
{
ParameterInfo[] ctorParams = ctors[i].GetParameters();
if (bestCtor == null || ctorParams.Length > bestParams.Length)
{
bestCtor = ctors[i];
bestParams = ctorParams;
}
}
if (bestCtor == null) throw new InvalidOperationException(
"Cannot create - no constructor");
object[] args = new object[bestParams.Length];
for (int i = 0; i < bestParams.Length; i++)
{
args[i] = propertyBag[bestParams[i].Name];
propertyBag.Remove(bestParams[i].Name);
}
object obj = bestCtor.Invoke(args);
// TODO: if we wanted, we could apply any unused keys in propertyBag
// at this point via properties
return obj;
}