How to convert a String to its equivalent LINQ Expression Tree?

StackOverflow https://stackoverflow.com/questions/821365

Question

This is a simplified version of the original problem.

I have a class called Person:

public class Person {
  public string Name { get; set; }
  public int Age { get; set; }
  public int Weight { get; set; }
  public DateTime FavouriteDay { get; set; }
}

...and lets say an instance:

var bob = new Person {
  Name = "Bob",
  Age = 30,
  Weight = 213,
  FavouriteDay = '1/1/2000'
}

I would like to write the following as a string in my favourite text editor....

(Person.Age > 3 AND Person.Weight > 50) OR Person.Age < 3

I would like to take this string and my object instance and evaluate a TRUE or FALSE - i.e. evaluating a Func<Person, bool> on the object instance.

Here are my current thoughts:

  1. Implement a basic grammar in ANTLR to support basic Comparison and Logical Operators. I am thinking of copying the Visual Basic precedence and some of the featureset here: http://msdn.microsoft.com/en-us/library/fw84t893(VS.80).aspx
  2. Have ANTLR create a suitable AST from a provided string.
  3. Walk the AST and use the Predicate Builder framework to dynamically create the Func<Person, bool>
  4. Evaluate the predicate against an instance of Person as required

My question is have I totally overbaked this? any alternatives?

EDIT: Chosen Solution

I decided to use the Dynamic Linq Library, specifically the Dynamic Query class provided in the LINQSamples.

Code below:

using System;
using System.Linq.Expressions;
using System.Linq.Dynamic;

namespace ExpressionParser
{
  class Program
  {
    public class Person
    {
      public string Name { get; set; }
      public int Age { get; set; }
      public int Weight { get; set; }
      public DateTime FavouriteDay { get; set; }
    }

    static void Main()
    {
      const string exp = @"(Person.Age > 3 AND Person.Weight > 50) OR Person.Age < 3";
      var p = Expression.Parameter(typeof(Person), "Person");
      var e = System.Linq.Dynamic.DynamicExpression.ParseLambda(new[] { p }, null, exp);
      var bob = new Person
      {
        Name = "Bob",
        Age = 30,
        Weight = 213,
        FavouriteDay = new DateTime(2000,1,1)
      };

      var result = e.Compile().DynamicInvoke(bob);
      Console.WriteLine(result);
      Console.ReadKey();
    }
  }
}

Result is of type System.Boolean, and in this instance is TRUE.

Many thanks to Marc Gravell.

Include System.Linq.Dynamic nuget package, documentation here

Was it helpful?

Solution

Would the dynamic linq library help here? In particular, I'm thinking as a Where clause. If necessary, put it inside a list/array just to call .Where(string) on it! i.e.

var people = new List<Person> { person };
int match = people.Where(filter).Any();

If not, writing a parser (using Expression under the hood) isn't hugely taxing - I wrote one similar (although I don't think I have the source) in my train commute just before xmas...

OTHER TIPS

Another such library is Flee

I did a quick comparison of Dynamic Linq Library and Flee and Flee was 10 times faster for the expression "(Name == \"Johan\" AND Salary > 500) OR (Name != \"Johan\" AND Salary > 300)"

This how you can write your code using Flee.

static void Main(string[] args)
{
  var context = new ExpressionContext();
  const string exp = @"(Person.Age > 3 AND Person.Weight > 50) OR Person.Age < 3";
  context.Variables.DefineVariable("Person", typeof(Person));
  var e = context.CompileDynamic(exp);

  var bob = new Person
  {
    Name = "Bob",
    Age = 30,
    Weight = 213,
    FavouriteDay = new DateTime(2000, 1, 1)
  };

  context.Variables["Person"] = bob;
  var result = e.Evaluate();
  Console.WriteLine(result);
  Console.ReadKey();
}

void Main()
{
    var testdata = new List<Ownr> {
        //new Ownr{Name = "abc", Qty = 20}, // uncomment this to see it getting filtered out
        new Ownr{Name = "abc", Qty = 2},
        new Ownr{Name = "abcd", Qty = 11},
        new Ownr{Name = "xyz", Qty = 40},
        new Ownr{Name = "ok", Qty = 5},
    };

    Expression<Func<Ownr, bool>> func = Extentions.strToFunc<Ownr>("Qty", "<=", "10");
    func = Extentions.strToFunc<Ownr>("Name", "==", "abc", func);

    var result = testdata.Where(func.ExpressionToFunc()).ToList();

    result.Dump();
}

public class Ownr
{
    public string Name { get; set; }
    public int Qty { get; set; }
}

public static class Extentions
{
    public static Expression<Func<T, bool>> strToFunc<T>(string propName, string opr, string value, Expression<Func<T, bool>> expr = null)
    {
        Expression<Func<T, bool>> func = null;
        try
        {
            var type = typeof(T);
            var prop = type.GetProperty(propName);
            ParameterExpression tpe = Expression.Parameter(typeof(T));
            Expression left = Expression.Property(tpe, prop);
            Expression right = Expression.Convert(ToExprConstant(prop, value), prop.PropertyType);
            Expression<Func<T, bool>> innerExpr = Expression.Lambda<Func<T, bool>>(ApplyFilter(opr, left, right), tpe);
            if (expr != null)
                innerExpr = innerExpr.And(expr);
            func = innerExpr;
        }
        catch (Exception ex)
        {
            ex.Dump();
        }

        return func;
    }
    private static Expression ToExprConstant(PropertyInfo prop, string value)
    {
        object val = null;

        try
        {
            switch (prop.Name)
            {
                case "System.Guid":
                    val = Guid.NewGuid();
                    break;
                default:
                    {
                        val = Convert.ChangeType(value, prop.PropertyType);
                        break;
                    }
            }
        }
        catch (Exception ex)
        {
            ex.Dump();
        }

        return Expression.Constant(val);
    }
    private static BinaryExpression ApplyFilter(string opr, Expression left, Expression right)
    {
        BinaryExpression InnerLambda = null;
        switch (opr)
        {
            case "==":
            case "=":
                InnerLambda = Expression.Equal(left, right);
                break;
            case "<":
                InnerLambda = Expression.LessThan(left, right);
                break;
            case ">":
                InnerLambda = Expression.GreaterThan(left, right);
                break;
            case ">=":
                InnerLambda = Expression.GreaterThanOrEqual(left, right);
                break;
            case "<=":
                InnerLambda = Expression.LessThanOrEqual(left, right);
                break;
            case "!=":
                InnerLambda = Expression.NotEqual(left, right);
                break;
            case "&&":
                InnerLambda = Expression.And(left, right);
                break;
            case "||":
                InnerLambda = Expression.Or(left, right);
                break;
        }
        return InnerLambda;
    }

    public static Expression<Func<T, TResult>> And<T, TResult>(this Expression<Func<T, TResult>> expr1, Expression<Func<T, TResult>> expr2)
    {
        var invokedExpr = Expression.Invoke(expr2, expr1.Parameters.Cast<Expression>());
        return Expression.Lambda<Func<T, TResult>>(Expression.AndAlso(expr1.Body, invokedExpr), expr1.Parameters);
    }

    public static Func<T, TResult> ExpressionToFunc<T, TResult>(this Expression<Func<T, TResult>> expr)
    {
        var res = expr.Compile();
        return res;
    }
}

LinqPad has the Dump() method

You might take a look at the DLR. It allows you to evaluate and execute scripts inside .NET 2.0 application. Here's a sample with IronRuby:

using System;
using IronRuby;
using IronRuby.Runtime;
using Microsoft.Scripting.Hosting;

class App
{
    static void Main()
    {
        var setup = new ScriptRuntimeSetup();
        setup.LanguageSetups.Add(
            new LanguageSetup(
                typeof(RubyContext).AssemblyQualifiedName,
                "IronRuby",
                new[] { "IronRuby" },
                new[] { ".rb" }
            )
        );
        var runtime = new ScriptRuntime(setup);
        var engine = runtime.GetEngine("IronRuby");
        var ec = Ruby.GetExecutionContext(runtime);
        ec.DefineGlobalVariable("bob", new Person
        {
            Name = "Bob",
            Age = 30,
            Weight = 213,
            FavouriteDay = "1/1/2000"
        });
        var eval = engine.Execute<bool>(
            "return ($bob.Age > 3 && $bob.Weight > 50) || $bob.Age < 3"
        );
        Console.WriteLine(eval);

    }
}

public class Person
{
    public string Name { get; set; }
    public int Age { get; set; }
    public int Weight { get; set; }
    public string FavouriteDay { get; set; }
}

Of course this technique is based on runtime evaluation and code cannot be verified at compile time.

Here is an example of a Scala DSL based parser combinator for parsing and evaluating of arithmetic expressions.

import scala.util.parsing.combinator._
/** 
* @author Nicolae Caralicea
* @version 1.0, 04/01/2013
*/
class Arithm extends JavaTokenParsers {
  def expr: Parser[List[String]] = term ~ rep(addTerm | minusTerm) ^^
    { case termValue ~ repValue => termValue ::: repValue.flatten }

  def addTerm: Parser[List[String]] = "+" ~ term ^^
    { case "+" ~ termValue => termValue ::: List("+") }

  def minusTerm: Parser[List[String]] = "-" ~ term ^^
    { case "-" ~ termValue => termValue ::: List("-") }

  def term: Parser[List[String]] = factor ~ rep(multiplyFactor | divideFactor) ^^
    {
      case factorValue1 ~ repfactor => factorValue1 ::: repfactor.flatten
    }

  def multiplyFactor: Parser[List[String]] = "*" ~ factor ^^
    { case "*" ~ factorValue => factorValue ::: List("*") }

  def divideFactor: Parser[List[String]] = "/" ~ factor ^^
    { case "/" ~ factorValue => factorValue ::: List("/") }

  def factor: Parser[List[String]] = floatingPointConstant | parantExpr

  def floatingPointConstant: Parser[List[String]] = floatingPointNumber ^^
    {
      case value => List[String](value)
    }

  def parantExpr: Parser[List[String]] = "(" ~ expr ~ ")" ^^
    {
      case "(" ~ exprValue ~ ")" => exprValue
    }

  def evaluateExpr(expression: String): Double = {
    val parseRes = parseAll(expr, expression)
    if (parseRes.successful) evaluatePostfix(parseRes.get)
    else throw new RuntimeException(parseRes.toString())
  }
  private def evaluatePostfix(postfixExpressionList: List[String]): Double = {
    import scala.collection.immutable.Stack

    def multiply(a: Double, b: Double) = a * b
    def divide(a: Double, b: Double) = a / b
    def add(a: Double, b: Double) = a + b
    def subtract(a: Double, b: Double) = a - b

    def executeOpOnStack(stack: Stack[Any], operation: (Double, Double) => Double): (Stack[Any], Double) = {
      val el1 = stack.top
      val updatedStack1 = stack.pop
      val el2 = updatedStack1.top
      val updatedStack2 = updatedStack1.pop
      val value = operation(el2.toString.toDouble, el1.toString.toDouble)
      (updatedStack2.push(operation(el2.toString.toDouble, el1.toString.toDouble)), value)
    }
    val initial: (Stack[Any], Double) = (Stack(), null.asInstanceOf[Double])
    val res = postfixExpressionList.foldLeft(initial)((computed, item) =>
      item match {
        case "*" => executeOpOnStack(computed._1, multiply)
        case "/" => executeOpOnStack(computed._1, divide)
        case "+" => executeOpOnStack(computed._1, add)
        case "-" => executeOpOnStack(computed._1, subtract)
        case other => (computed._1.push(other), computed._2)
      })
    res._2
  }
}

object TestArithmDSL {
  def main(args: Array[String]): Unit = {
    val arithm = new Arithm
    val actual = arithm.evaluateExpr("(12 + 4 * 6) * ((2 + 3 * ( 4 + 2 ) ) * ( 5 + 12 ))")
    val expected: Double = (12 + 4 * 6) * ((2 + 3 * ( 4 + 2 ) ) * ( 5 + 12 ))
    assert(actual == expected)
  }
}

The equivalent expression tree or parse tree of the provided arithmetic expression would be of the Parser[List[String]] type.

More details are at the following link:

http://nicolaecaralicea.blogspot.ca/2013/04/scala-dsl-for-parsing-and-evaluating-of.html

In addition to Dynamic Linq Library (which builds strongly typed expression and requires strongly typed variables) I recommend better alternative: linq parser that part of NReco Commons Library (open source). It aligns all types and performs all invocations at runtime and behaves like dynamic language:

var lambdaParser = new NReco.LambdaParser();
var varContext = new Dictionary<string,object>();
varContext["one"] = 1M;
varContext["two"] = "2";

Console.WriteLine( lambdaParser.Eval("two>one && 0<one ? (1+8)/3+1*two : 0", varContext) ); // --> 5
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top