Related
I am attempting to implement a method like:
(Func<T> getFn, Action<T> setFn) MakePair<T>(T initialVal) {
}
It will return two runtime generated lambdas that get and set a dynamically created variable using Expression trees to create the code.
My current solution is to dynamically create an array of the type with one element, and reference that:
(Func<T> getFn, Action<T> setFn) MakePair<T>(T initialVal) {
var dynvar = Array.CreateInstance(typeof(T), 1);
Expression<Func<Array>> f = () => dynvar;
var dynref = Expression.Convert(f.Body, typeof(T).MakeArrayType());
var e0 = Expression.Constant(0);
var getBody = Expression.ArrayIndex(dynref, e0);
var setParam = Expression.Parameter(typeof(T));
var setBody = Expression.Assign(Expression.ArrayAccess(dynref, e0), setParam);
var getFn = Expression.Lambda<Func<T>>(getBody).Compile();
var setFn = Expression.Lambda<Action<T>>(setBody, setParam).Compile();
return (getFn, setFn);
}
Is there a better way to create what may be a value type variable at runtime that can be read/written to than using an array?
Is there a better way to reference the runtime created array other than using a lambda to create the (field?) reference for use in the ArrayIndex/ArrayAccess method calls?
Excessive Background Info
For those that wonder, ultimately this came up in an attempt to create something like Perl auto-virification of lvalues for Perl hashes.
Imagine you have a List<T> with duplicate elements and want to create a Dictionary<T,int> that allows you to look up the count for each unique T in the list. You can use a few lines of code to count (in this case T is int):
var countDict = new Dictionary<int, int>();
foreach (var n in testarray) {
countDict.TryGetValue(n, out int c);
countDict[n] = c + 1;
}
But I want to do this with LINQ, and I want to avoid double-indexing countDict (interestingly, ConcurrentDictionary has AddOrUpdate for this purpose) so I use Aggregate:
var countDict = testarray.Aggregate(new Dictionary<int,int>(), (d, n) => { ++d[n]; return d; });
But this has a couple of issues. First, Dictionary won't create a value for a missing value, so you need a new type of Dictionary that auto-creates missing values using e.g. a seed lambda:
var countDict = testarray.Aggregate(new SeedDictionary<int, Ref<int>>(() => Ref.Of(() => 0)), (d, n) => { var r = d[n]; ++r.Value; return d; });
But you still have the lvalue problem, so you replace the plain int counter with a Ref class. Unfortunately, C# can't create a C++ first class Ref class, but using one based around auto-creating a setter lambda from a getter lambda (using expression trees) is close enough. (Unfortunately C# still won't accept ++d[n].Value; even though it should be valid, so you have to create a temporary.)
But now you have the problem of creating multiple runtime integer variables to hold the counts. I extended the Ref<> class to take a lambda that returns a constant (ConstantExpression) and create a runtime variable and build a getter and setter with the constant being the initial value.
I agree with some of the question commenters that expression trees seem unnecessary, so here is a simple implementation of the shown API without them:
struct Box<T> {
public T Value;
}
(Func<T> getFn, Action<T> setFn) MakePair<T>(T initialVal) {
var box = new Box<T> { Value = initialVal };
return (() => box.Value, v => box.Value = v);
}
As an answer to the stated question (how to define dynref without a lambda), then, is there something wrong with the following modifications to dynvar and dynref?
var dynvar = new T[] { initialVal };
var dynref = Expression.Constant(dynvar);
I'm trying to call a function in a dynamic linq select statement, but im getting error:
No property or field 'A' exists in type 'Tuple2'
Example code:
void Main()
{
var a = new Tuple<int, int>(1,1);
var b = new[]{ a };
var q = b.AsQueryable().Select("A.Test(it.Item1)");
q.Dump();
}
public static class A
{
public static int Test(int i)
{
return i++;
}
}
How should I change my code to get this working?
If I call built in function Convert.ToInt32 for example it works fine.
var q = b.AsQueryable().Select("Convert.ToInt32(it.Item1)");
Also how do I cast a property using dynamic linq?
var q = b.AsQueryable().Select("((float)it.Item1)");
I'll say that the dynamic-linq isn't "strong enough" to do these things. It looks for methods only in the given objects and some special classes: Math, Convert, the various base types (int, float, string, ...), Guid, Timespan, DateTime
The list of these types is clearly visible if you use ilspy/reflector on the file. They are in System.Linq.Dynamic.ExpressionParser.predefinedTypes .
Now, clearly I could be wrong, but this works: .Select("Guid.NewGuid().ToString()").Cast<string>().ToArray()
showing that it's quite probable that that is the "correct" list.
There is an article here on how to modify Dynamic LINQ if you are interested http://www.krizzcode.com/2012/01/extending-dynamiclinq-language.html
Now, an intelligent man would take the source of dynamic linq and simply expand that array... But here there aren't intelligent men... There are only programmers that want blood! Blood but especially innards!
var type = typeof(DynamicQueryable).Assembly.GetType("System.Linq.Dynamic.ExpressionParser");
FieldInfo field = type.GetField("predefinedTypes", BindingFlags.Static | BindingFlags.NonPublic);
Type[] predefinedTypes = (Type[])field.GetValue(null);
Array.Resize(ref predefinedTypes, predefinedTypes.Length + 1);
predefinedTypes[predefinedTypes.Length - 1] = typeof(A); // Your type
field.SetValue(null, predefinedTypes);
Do this (with the types you want) BEFORE the first call to Dynamic Linq (because after the first call the methods/properties of these types are cached)
Explanation: we use reflection to add our object(s) to this "special list".
I know there is already an accepted answer on this but it did not work for me. I am using Dynamic Linq 1.1.4. I wanted to do a query like this
$.GetNewestRisk() == null
Where GetNewestRisk() is a public method on the object represented by $. I kept getting this error "Error running query, Methods on type 'Patient' are not accessible (at index 2)".
I found in the source code there is a GlobalConfig object that allows a custom provider to be assigned which will hold all of the types you may want to work with. Here is the source code for the custom provider:
public class CustomTypeProvider: IDynamicLinkCustomTypeProvider
{
public HashSet<Type> GetCustomTypes()
{
HashSet<Type> types = new HashSet<Type>();
types.Add(typeof(Patient));
types.Add(typeof(RiskFactorResult));
types.Add(typeof(PatientLabResult));
types.Add(typeof(PatientVital));
return types;
}
}
Here is how I am using it:
System.Linq.Dynamic.GlobalConfig.CustomTypeProvider = new CustomType();
After making this call I am able to call methods on the objects inside of the expression.
#xanatos answer doesn't work for .Net Core version. So I've found something similar related by #Kent on the System.Dynamic.Linq.Core tests DynamicExpressionParserTests written by the library's author himself.
The given TestCustomTypeProviderClass allows you to use the DynamicLinqType class annotation which is pretty usefull for this problem.
To answer to question, you then just needed to defined the class (ensure to annotate with DynamicLinqType) :
[DynamicLinqType]
public static class A
{
public static int Test(int i)
{
return i++;
}
}
Add a customTypeProvider as mentioned above :
private class TestCustomTypeProvider : AbstractDynamicLinqCustomTypeProvider, IDynamicLinkCustomTypeProvider
{
private HashSet<Type> _customTypes;
public virtual HashSet<Type> GetCustomTypes()
{
if (_customTypes != null)
{
return _customTypes;
}
_customTypes = new HashSet<Type>(FindTypesMarkedWithDynamicLinqTypeAttribute(new[] { GetType().GetTypeInfo().Assembly }));
return _customTypes;
}
}
and use a ParsingConfig with the configurable Select to call it :
var config = new ParsingConfig
{
CustomTypeProvider = new TestCustomTypeProvider()
};
var q = b.AsQueryable().Select(config, "A.Test(it.Item1)");
#Armand has put together a brilliant solution for this issue, and being the only solution I was able to find regarding this I want to add to it for anyone who tries the same approach.
The class that is marked with...
[DynamicLinqType]
... must be taken into consideration when you run the following line:
FindTypesMarkedWithDynamicLinqTypeAttribute(new[] { GetType().GetTypeInfo().Assembly })
In the solution provided above, this assumes the class that contains the function to be evaluated is on the same class the code currently resides in. If the methods are to be used outside of said class, the assembly will need to change.
FindTypesMarkedWithDynamicLinqTypeAttribute(new[] { typeof(AnotherClassName).Assembly })
Nothing changes from the solution above, this is just for clarification for anyone attempting to use it.
As regards the current version (1.2.19) of Dynamic LINQ, you will probably get another exception:
System.Linq.Dynamic.Core.Exceptions.ParseException : Enum value 'Test' is not defined in enum type 'A'
To make DLINQ know your type 'A', you have two options:
Set up parsing config with your own custom types provider where you directly specify the type 'A'.
Mark your type with the attribute [DynamicLinqType]. If that type is loaded into the current domain (that's the usual case), you don't have to do anything more since the default custom type provider already scans the current AppDomain for types marked with [DynamicLinqType]. And only if that's not the case, i.e. your type is not loaded into the current domain, you have to do something like in that answer.
What if you would like to use both approaches - the first for type 'A' and the second for type 'B'? In that case, you just have to "merge" your type 'A' with the default provider types:
public class DynamicLinqTests
{
[Test]
public void Test()
{
var a = new Tuple<int, int>(1, 1);
var b = new[] { a };
var parsingConfig = new ParsingConfig
{
ResolveTypesBySimpleName = true,
CustomTypeProvider = new TestCustomTypesProvider()
};
var queryWithA = b.AsQueryable().Select(parsingConfig, "A.Test(it.Item1)");
queryWithA.ToDynamicList();
var queryWithB = b.AsQueryable().Select(parsingConfig, "B.Test(it.Item1)");
queryWithB.ToDynamicList();
}
public static class A
{
public static int Test(int i)
{
return i++;
}
}
[DynamicLinqType]
public static class B
{
public static int Test(int i)
{
return i++;
}
}
public class TestCustomTypesProvider : DefaultDynamicLinqCustomTypeProvider
{
public override HashSet<Type> GetCustomTypes()
{
var customTypes = base.GetCustomTypes();
customTypes.Add(typeof(A));
return customTypes;
}
}
}
I may be confused but your syntax whereby you are using a string in your Selects doesn't compile for me. The following syntax works:
var q = b.AsQueryable().Select(it => A.Test(it.Item1));
var b = new[]{ a };
The above array is don't know what type of array , and it's not type safe ?
Your values are assigned in variant data type so it's not integer value (I think string value) ,when you get this values in your query must need to convert.toint32() because your class parameter data type is integer
Please try it
var b = new **int**[]{ a };
instead of var b = new[]{ a };
The important hint is here (in bold):
No property or field 'xxx' exists in **type** 'xxx'
And Please look this for previous discussion :
Dynamic Linq - no property or field exists in type 'datarow'
The following works for me:
var a = new Tuple<int, int>(1, 1);
var b = new[] { a };
var q = b.AsQueryable().Select(it=>A.Test(it.Item1));
var q1 = b.AsQueryable().Select(it => Convert.ToInt32(it.Item1));
var q2 = b.AsQueryable().Select(it => (float) it.Item1);
I have some simple C# types that are used in an embedded IronPython script:
//simple type
public class Foo
{
public int a;
public int b;
}
var engine = IronPython.Hosting.Python.CreateEngine();
dynamic scope = engine.CreateScope();
scope.foo = new Foo{ a = 1, b = 2 };
engine.Execute( "print( foo.a )", scope );
I'd like to add some functionality to Foo but I cannot modify it's code. Also I'd rather not derive from it nor use extension methods, but use delegates instead. Ideally I'd like to write something like
scope.AddMemberFunction( scope.foo, "Fun",
new Func<Foo,int>( f => return f.a + f.b ) );
and then use it in the Python script directly:
print( foo.Fun() )
I thought this was exactly what could be done using the Operations but this raises an exception saying 'Foo' object has no attribute 'Fun'
engine.Operations.SetMember( scope.foo, "Fun",
new Func<Foo, int>( f => f.a + f.b ) );
Next I tried the Python way of things (say Foo is in a namespace IronPythonApp that is in and it's assembly is added to the engine.Runtime):
import IronPythonApp
def Fun( self ):
return self.a + self.b
IronPythonApp.Foo.Fun = Fun
but this gives a similar exception: can't set attributes of built-in/extension type 'Foo'
Is there a way to modify the Python class definition that ironPython internally generates?
Update
I explored some of the options as by Jeff Hardy's answer. Here are two ways that can both be made pretty scalable (just showing the straightforward way here)
ExpandoObject!
var foo = new Foo { a = 1, b = 2 };
dynamic exp = new ExpandoObject();
exp.Fun = new Func<int>( () => foo.a + foo.b );
exp.a = foo.a; //automating these is not a big deal
exp.b = foo.b; //
//now just add exp to the scope and done.
Using SetMember after all
scope.origFoo = new Foo { a = 1, b = 2 };
//the only variable thing in this string is 'origFoo'
//so it's no big deal scaling this
engine.Execute( #"
class GenericWrapper( object ) :
def __init__( self, foo ):
self.foo = foo
def __getattr__( self, name ) :
return getattr( self.foo, name )
foo = GenericWrapper( origFoo )", scope );
//ha, now scope contains something that we can mess with after all
engine.Operations.SetMember( scope.foo, "Fun",
new Func<int>( () => scope.foo.a + scope.foo.b ) );
engine.Execute( "print( foo.Fun() )", scope );
Short answer: nope, you can't modify a .NET class. By default, they don't have the necessary dynamic hooks to add members.
Your best bet is to wrap the class and add the members you want; Python makes this easy:
class FooWrapper(object):
def __init__(self, foo):
self.foo = foo
def __getattr__(self, name):
return getattr(self.foo, name)
def mymethod(self):
...
The __getattr__ special method is only called for attributes that are not part of normal attribute lookup, so mymethod() will be looked up normally but anything else will be forwarded to the underlying object.
If you need to do it on the C# side, you can achieve the same thing by subclassing DynamicObject and overloading the Try*Member functions.
This question already has answers here:
Closed 10 years ago.
Possible Duplicate:
How can I pass an anonymous type to a method?
I have the following LINQ Statement, whose output has to be processed in another method:
var data = from lines in File.ReadAllLines(TrainingDataFile)
.Skip(ContainsHeader ? 1 : 0)
let f = lines.Split(new[] { FieldSeparator }).ToList<String>()
let target = f[TargetVariablePositionZeroBased]
select new { F=f, T=target };
What should be the datatype of the parameter in the method that will take this data?
You can not return the anonymous data types from a method. You can define a class and return object of that class from query and pass it to target method.
public class SomeClass
{
public string F {get; set;}
public string T {get; set;}
}
var data = from lines in File.ReadAllLines(TrainingDataFile)
.Skip(ContainsHeader ? 1 : 0)
let f = lines.Split(new[] { FieldSeparator }).ToList<String>()
let target = f[TargetVariablePositionZeroBased]
select new SomeClass { F=f, T=target };
You can pass the query result IEnumerable<SomeClass> to method as parameter.
public void MethodToCall(IEnumerable<SomeClass> someClass)
{
}
To call the method by passing the query result (IEnumerable<SomeClass>) that is stored in data in this sample code
MethodToCall(data);
You can't very easily pass anonymous types around. You can either create a class, or since your data has only two properties, use a Tuple:
select new Tuple<List<string>, string> (f, target);
If I have the data types correct, then the data type of the parameter would be:
IEnumerable<Tuple<List<string>, string>>
and you would reference F and T using the Tuple properties Item1 and Item2.
1) Just to pass the result of the query, make your function generic, that will do:
var data = from lines in File.ReadAllLines(TrainingDataFile)
.Skip(ContainsHeader ? 1 : 0)
let f = lines.Split(new[] { FieldSeparator }).ToList<String>()
let target = f[TargetVariablePositionZeroBased]
select new { F=f, T=target };
SomeMethod(data);
public void SomeMethod<T>(IEnumerable<T> enumerable)
{
// ^^choose the return type..
}
Simple. If the processing inside the method is something so simple this will do. But you won't be able to access properties F and T inside the method.
To do so:
2) You can use the "cast by example" trick shown here by Eric. To quote him:
We use method type inference and local variable type inference to tell
the compiler "these two things are the same type". This lets you
export an anonymous type as object and cast it back to anonymous type.
...the trick only works if the example and the source objects were
created in code in the same assembly; two "identical" anonymous types
in two different assemblies do not unify to be the same type.
SomeMethod(data);
public void SomeMethod(IEnumerable<object> enumerable)
{
var template = new { F = new List<string>(), T = string.Empty };
foreach (var item in enumerable)
{
var anonymousType = item.CastToTypeOf(template);
//print string.Join(", ", anonymousType.F) + " - " + anonymousType.T //compiles
//or whatever
}
}
//a more generic name perhaps is 'CastToTypeOf' as an extension method
public static T CastToTypeOf<T>(this object source, T example) where T : class
{
return (T)source;
}
The catch here is that SomeMethod now is tailor made for your anonymous type, since you're specifying a specific type inside the method, so its better to not make the function generic (though you can do) and to give a suitable name for the function.
3) If function is just for your unique type now, I would better have them all wrapped in a single method and not pass at all - no hassle! :)
4) Or you can delegate the action to be done on your anonymous type. So method signature would be like:
SomeMethod(data, d => print string.Join(", ", d.F) + " - " + d.T);
public void SomeMethod<T>(IEnumerable<T> enumerable, Action<T> actor)
{
foreach (var item in enumerable)
actor(item);
}
If it matters you can have Func delegate as well by having one more type argument.
5) Rely on fiddly reflection to get the properties from your anonymous type otherwise.
6) Use dynamic keyword on method argument and now you have dynamic typing. Both the above doesnt give you benefits of static typing.
7) You will be better off having a separate class that holds F and T. And that the best of all. But ask yourself do they together represent something as an entity?
8) If not, just pass an IEnumerable<Tuple> or IDictionary depending on what matters.
It all depends on what/how you want to achieve with the method. Personally, I would go for the approach 2 in a hobby project (for the fun involved), but in production code 3, 4, 7, 8 depending on the context.
Is it possible to evaluate the following in C# at runtime
I have a class that contains 3 properties (Field,Operator,Value)
rule.Field;
rule.Operator;
rule.Value;
this is my rule class...
Now I have a loop
foreach(item in items)
{
// here I want to create a dynamic expression to evaluate at runtime
// something like
if (item.[rule.field] [rule.operator] [rule.value])
{ do work }
}
I just don't know the syntax, or if its possible in C#, I know in JS its possible but that's not a compiled language.
Update
Essentially I want a way to eval(stringCode) or a better more supported way.
No, C# doesn't support anything like this directly.
The closest options are:
Create a full valid C# program and dynamically compile it with CSharpCodeProvider.
Build an expression tree, compile and execute it
Perform the evaluation yourself (this may actually be easiest, depending on your operators etc)
Disclaimer: I'm the owner of the project Eval Expression.NET
This library is close to being the JS Eval equivalent. You can almost evaluate and compile all the C# language.
Here is a simple example using your question, but the library goes way beyond this simple scenario.
int field = 2;
int value = 1;
string binaryOperator = ">";
string formula = "x " + binaryOperator + " y";
// For single evaluation
var value1 = Eval.Execute<bool>(formula, new { x = field, y = value });
// For many evaluation
var compiled = Eval.Compile<Func<int, int, bool>>(formula, "x", "y");
var value2 = compiled(field, value);
EDIT Answer comment:
Proprietary library to do simple evaluation? No, thanks
This library does not support only simple evaluation but almost all the C# languages. Allowing you to add dynamically a method, use async, linq, loop, etc., which is more than "to do simple evaluation"
The closest options solution provided by Jon Skeet are great but will surely take several days of development and testing to support all cases, depending on the operators. Surely this library helps some developers, but in some other scenarios, like yours, it could be done without it.
I'm not entirely sure what you are saying. Can you try clarifying it a bit?
Are you wanting to to take a string expression and evaluate it at runtime in C#? If so the answer is no. C# does not support such types of dynamic evaluation.
You'd have to either use the CodeDOM libraries or create an Expression tree, compile it, and execute it. I think building up the expression tree is the best option.
Of course you could put in a switch statement on your operator, which is not bad because there is a limited number of operators you could use anyways.
Here's a way to do this with expression trees (written in LINQPad):
void Main()
{
var programmers = new List<Programmer>{
new Programmer { Name = "Turing", Number = Math.E},
new Programmer { Name = "Babbage", Number = Math.PI},
new Programmer { Name = "Lovelace", Number = Math.E}};
var rule0 = new Rule<string>() { Field = "Name", Operator = BinaryExpression.Equal, Value = "Turing" };
var rule1 = new Rule<double>() { Field = "Number", Operator = BinaryExpression.GreaterThan, Value = 2.719 };
var matched0 = RunRule<Programmer, string>(programmers, rule0);
matched0.Dump();
var matched1 = RunRule<Programmer, double>(programmers, rule1);
matched1.Dump();
var matchedBoth = matched0.Intersect(matched1);
matchedBoth.Dump();
var matchedEither = matched0.Union(matched1);
matchedEither.Dump();
}
public IEnumerable<T> RunRule<T, V>(IEnumerable<T> foos, Rule<V> rule) {
var fieldParam = Expression.Parameter(typeof(T), "f");
var fieldProp = Expression.Property (fieldParam, rule.Field);
var valueParam = Expression.Parameter(typeof(V), "v");
BinaryExpression binaryExpr = rule.Operator(fieldProp, valueParam);
var lambda = Expression.Lambda<Func<T, V, bool>>(binaryExpr, fieldParam, valueParam);
var func = lambda.Compile();
foreach(var foo in foos) {
var result = func(foo, rule.Value);
if(result)
yield return foo;
}
}
public class Rule<T> {
public string Field { get; set; }
public Func<Expression, Expression, BinaryExpression> Operator { get; set; }
public T Value { get; set; }
}
public class Programmer {
public string Name { get; set; }
public double Number { get; set; }
}
A better design for you would be for your rule to apply the test itself (or to an arbitrary value)
By doing this with Func instances you will get the most flexibility, like so:
IEnumerable<Func<T,bool> tests; // defined somehow at runtime
foreach (var item in items)
{
foreach (var test in tests)
{
if (test(item))
{
//do work with item
}
}
}
then your specific test would be something like this for strong type checking at compile time:
public Func<T,bool> FooEqualsX<T,V>(V x)
{
return t => EqualityComparer<V>.Default.Equals(t.Foo, x);
}
For a reflective form
public Func<T,bool> MakeTest<T,V>(string name, string op, V value)
{
Func<T,V> getter;
var f = typeof(T).GetField(name);
if (f != null)
{
if (!typeof(V).IsAssignableFrom(f.FieldType))
throw new ArgumentException(name +" incompatible with "+ typeof(V));
getter= x => (V)f.GetValue(x);
}
else
{
var p = typeof(T).GetProperty(name);
if (p == null)
throw new ArgumentException("No "+ name +" on "+ typeof(T));
if (!typeof(V).IsAssignableFrom(p.PropertyType))
throw new ArgumentException(name +" incompatible with "+ typeof(V));
getter= x => (V)p.GetValue(x, null);
}
switch (op)
{
case "==":
return t => EqualityComparer<V>.Default.Equals(getter(t), value);
case "!=":
return t => !EqualityComparer<V>.Default.Equals(getter(t), value);
case ">":
return t => Comparer<V>.Default.Compare(getter(t), value) > 0;
// fill in the banks as you need to
default:
throw new ArgumentException("unrecognised operator '"+ op +"'");
}
}
If you wanted to be really introspective and handle any literal without knowing at compile time you could use the CSharpCodeProvider to compile a function assuming something like:
public static bool Check(T t)
{
// your code inserted here
}
This is of course a massive security hole so whoever can supply code for this must be fully trusted. Here is a somewhat limited implementation for your specific needs (no sanity checking at all)
private Func<T,bool> Make<T>(string name, string op, string value)
{
var foo = new Microsoft.CSharp.CSharpCodeProvider()
.CompileAssemblyFromSource(
new CompilerParameters(),
new[] { "public class Foo { public static bool Eval("+
typeof(T).FullName +" t) { return t."+
name +" "+ op +" "+ value
+"; } }" }).CompiledAssembly.GetType("Foo");
return t => (bool)foo.InvokeMember("Eval",
BindingFlags.Static | BindingFlags.Public | BindingFlags.InvokeMethod ,
null, null, new object[] { t });
}
// use like so:
var f = Make<string>("Length", ">", "2");
For this to work with arbitrary types you would have to do a bit more reflection to find the target assembly for the type to reference it in the compiler parameters.
private bool Eval(object item, string name, string op, string value)
{
var foo = new Microsoft.CSharp.CSharpCodeProvider()
.CompileAssemblyFromSource(
new CompilerParameters(),
new[] { "public class Foo { public static bool Eval("+
item.GetType().FullName +" t) "+
"{ return t."+ name +" "+ op +" "+ value +"; } }"
}).CompiledAssembly.GetType("Foo");
return (bool)foo.InvokeMember("Eval",
BindingFlags.Static | BindingFlags.Public | BindingFlags.InvokeMethod ,
null, null, new object[] { item });
}
All the above code is simply a proof of concept, it lacks sanity checking and has serious performance issues.
If you wanted to be even fancier you could use Reflection.Emit with DynamicMethod instances to do it (using proper operators rather than the default comparer instances) but this would require complex handling for types with overridden operators.
By making your check code highly generic you may include more tests in future as you need to. Essentially isolate the part of your code that cares only about a function from t -> true/false from the code that supplies these functions.
CSharpCodeProvider; switch statements that pick the proper different "operators"; the DLR... they are all ways you could do this; but they seem weird solutions to me.
How about just using delegates?
Assuming your Field and Value are numbers, declare something like this:
delegate bool MyOperationDelegate(decimal left, decimal right);
...
class Rule {
decimal Field;
decimal Value;
MyOperationDelegate Operator;
}
Now you can define your 'rule' as, for example, a bunch of lambdas:
Rule rule1 = new Rule;
rule1.Operation = (decimal l, decimal r) => { return l > r; };
rule1.Field = ...
You can make arrays of rules and apply them whichever way you wish.
IEnumerable<Rule> items = ...;
foreach(item in items)
{
if (item.Operator(item.Field, item.Value))
{ /* do work */ }
}
If Field and Values are not numbers, or the type depends on the specific rule, you can use object instead of decimal, and with a little bit of casting you can make it all work.
That's not a final design; it's just to give you some ideas (for example, you would likely have the class evaluate the delegate on its own via a Check() method or something).
You can retrieve the field by reflection. And then implement the operators as methods and uses reflection or some types of enum-delegate mapping to call the operators. The operators should have at least 2 parameters, the input value and the value you are using to test against with.
While it is true that you probably won't find an elegant way to evaluate full C# code on the fly without the use of dynamically compiling code (which is never pretty), you can almost certainly get your rules evaluated in short order using either the DLR (IronPython, IronRuby, etc) or an expression evaluator library that parses and executes a custom syntax. There is one, Script.NET, that provides a very similar syntax to C#.
Take a look here:Evaluating Expressions a Runtime in .NET(C#)
If you have the time / inclination to learn a little Python, then IronPython and the DLR will solve all your issues:
Extending your App with IronPython