22

I'm trying to build something like conditional queries to get only needed data from the underlying database.

Currently I have the following query (which works fine)

var eventData = dbContext.Event.Select(t => new { Address = true ? new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName } : new AnonymousEventGetAddress(), });

If I change it to

var includeAddress = true; // this will normally be passed as param var eventData = dbContext.Event.Select(t => new { Address = includeAddress ? new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName } : new AnonymousEventGetAddress(), });

I get the following error:

The type 'AnonymousEventGetAddress' appears in two structurally incompatible initializations within a single LINQ to Entities query. A type can be initialized in two places in the same query, but only if the same properties are set in both places and those properties are set in the same order.

What am I doing wrong here (as of with the true it's working) and how can this be fixed?

I know that changing the else-part to

new AnonymousEventGetAddress { AddressLine1 = null, CityName = null }

will work. But if I change the order of the properties then, this will also fail.

The class used is defined the following:

public class AnonymousEventGetAddress : BaseAnonymousObject<AnonymousEventGetAddress> { public string AddressLine1 { get; set; } public string CityName { get; set; } }

whereas BaseAnonymousObject<AnonymousEventGetAddress> is defined:

public abstract class BaseAnonymousObject<TAnonymous> where TAnonymous : BaseAnonymousObject<TAnonymous> { // this is used in case I have to return a list instead of a single anonymous object public static Expression<Func<IEnumerable<TAnonymous>>> Empty => () => new TAnonymous[]{}.AsEnumerable(); }
Improve this question
4
  • I would suggest that you don't do any initialisation in the query, leave the query to get what you need then cast that anonymous type back to something (DTO) and then go through that initialising everything. Remember you are working with IQueryable so Expression of Func so these will be compiled to LINQ, they aren't executed... Commented Aug 25, 2016 at 8:04
  • @CallumLinington I know at runtime, which data is needed. So I was looking for a way to select only what is really needed also from database to keep the performance good. On the other hand I use LINQKit to enable the usage of Expression<> which work pretty fine when passing true. Commented Aug 25, 2016 at 8:09
  • Create two Select expressions, so in one you are doing it if includeAddress or choose the other if otherwise. Don't try and compound it into the expression it will get treated as an expression... Commented Aug 25, 2016 at 8:10
  • It was just a simplified example. In case I have three optional include... this would end up in 8 different queries for each combination of them... So I'm still wondering what's different between passing includeAdress (fails) and true (working). Commented Aug 25, 2016 at 8:15

6 Answers 6

Reset to default
9

I don't know why EF has such requirement, but the important thing is that the requirement exists and we need to take it into account.

The first code works because true is a compile time constant, so the compiler is resolving it at compile time, ending up with one of the two expressions (basically removing the ternary operator). While in the second case it's a variable, thus the expression tree contains the original expression and fails at runtime due to aforementioned EF requirement.

A while ago I was trying to solve this and similar problems (to be honest, mainly for dynamic where filters) by implementing a custom method which is trying to resolve the bool variables, thus doing at runtime something similar to what does the compiler in the first case. Of course the code is experimental and not tested, but seem to handle properly such scenarios, so you can give it a try. The usage is quite simple:

var eventData = dbContext.Event.Select(t => new { Address = includeAddress ? new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName } : new AnonymousEventGetAddress(), }).ReduceConstPredicates();

And here is the helper method used:

public static partial class QueryableExtensions { public static IQueryable<T> ReduceConstPredicates<T>(this IQueryable<T> source) { var visitor = new ConstPredicateReducer(); var expression = visitor.Visit(source.Expression); if (expression != source.Expression) return source.Provider.CreateQuery<T>(expression); return source; } class ConstPredicateReducer : ExpressionVisitor { int evaluateConst; private ConstantExpression TryEvaluateConst(Expression node) { evaluateConst++; try { return Visit(node) as ConstantExpression; } finally { evaluateConst--; } } protected override Expression VisitConditional(ConditionalExpression node) { var testConst = TryEvaluateConst(node.Test); if (testConst != null) return Visit((bool)testConst.Value ? node.IfTrue : node.IfFalse); return base.VisitConditional(node); } protected override Expression VisitBinary(BinaryExpression node) { if (node.Type == typeof(bool)) { var leftConst = TryEvaluateConst(node.Left); var rightConst = TryEvaluateConst(node.Right); if (leftConst != null || rightConst != null) { if (node.NodeType == ExpressionType.AndAlso) { if (leftConst != null) return (bool)leftConst.Value ? (rightConst ?? Visit(node.Right)) : Expression.Constant(false); return (bool)rightConst.Value ? Visit(node.Left) : Expression.Constant(false); } else if (node.NodeType == ExpressionType.OrElse) { if (leftConst != null) return !(bool)leftConst.Value ? (rightConst ?? Visit(node.Right)) : Expression.Constant(true); return !(bool)rightConst.Value ? Visit(node.Left) : Expression.Constant(true); } else if (leftConst != null && rightConst != null) { var result = Expression.Lambda<Func<bool>>(Expression.MakeBinary(node.NodeType, leftConst, rightConst)).Compile().Invoke(); return Expression.Constant(result); } } } return base.VisitBinary(node); } protected override Expression VisitMethodCall(MethodCallExpression node) { if (evaluateConst > 0) { var objectConst = node.Object != null ? TryEvaluateConst(node.Object) : null; if (node.Object == null || objectConst != null) { var arguments = new object[node.Arguments.Count]; bool canEvaluate = true; for (int i = 0; i < arguments.Length; i++) { var argumentConst = TryEvaluateConst(node.Arguments[i]); if (canEvaluate = (argumentConst != null)) arguments[i] = argumentConst.Value; else break; } if (canEvaluate) { var result = node.Method.Invoke(objectConst != null ? objectConst.Value : null, arguments); return Expression.Constant(result, node.Type); } } } return base.VisitMethodCall(node); } protected override Expression VisitUnary(UnaryExpression node) { if (evaluateConst > 0 && (node.NodeType == ExpressionType.Convert || node.NodeType == ExpressionType.ConvertChecked)) { var operandConst = TryEvaluateConst(node.Operand); if (operandConst != null) { var result = Expression.Lambda(node.Update(operandConst)).Compile().DynamicInvoke(); return Expression.Constant(result, node.Type); } } return base.VisitUnary(node); } protected override Expression VisitMember(MemberExpression node) { object value; if (evaluateConst > 0 && TryGetValue(node, out value)) return Expression.Constant(value, node.Type); return base.VisitMember(node); } static bool TryGetValue(MemberExpression me, out object value) { object source = null; if (me.Expression != null) { if (me.Expression.NodeType == ExpressionType.Constant) source = ((ConstantExpression)me.Expression).Value; else if (me.Expression.NodeType != ExpressionType.MemberAccess || !TryGetValue((MemberExpression)me.Expression, out source)) { value = null; return false; } } if (me.Member is PropertyInfo) value = ((PropertyInfo)me.Member).GetValue(source); else value = ((FieldInfo)me.Member).GetValue(source); return true; } } }
Improve this answer
Sign up to request clarification or add additional context in comments.

3 Comments

Just tested your code. Unfortunately it's not working. Still the same error.
Darn, when I wrote it's experimental I didn't mean it should fail immediately:) Actually I've tested your scenario on my EF test environment and it worked. The problem is that my helper method code was different from the one that I posted in the links (let me not forget to update them later as well) and copy/pasted here. Now it's up to date. Besides that, hope you understand why the first case with constant true works while the variable true does not.
No problem. Will test your new code later. And yes. I understand now why true is working. Thanks for the clarification on that. Removing the else part makes totally sense and there will be no double initialization left which causes EF to crash.
9

For future readers, this SO duplicate (added one year later) was key to solving my problems:

The type appear in two structurally incompatible initializations within a single LINQ to Entities query

When you look at it, the error message is abundantly clear. Don't mess up the initialization order if you are instantiating an object more than once in the same Linq expression. For me, that was exactly what I was doing. Synchronizing the property initializations between the two instantiation calls had the compiler blowing sunshine again.

In this case:

new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName }

is different from 

new AnonymousEventGetAddress()

In OP query version 1 it is safe to say that the aberrant initialization in the else branch could never happen due to the the true conditional, why it was probably discarded, for version two that must not have happened, and we're left with two initialization orders, properties 1 and 2 versus no properties at all. This should do it:

includeAddress ? new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName } : new AnonymousEventGetAddress { AddressLine1 = null, CityName = null }
Improve this answer

1 Comment

This solved my issue where I was using a common DTO to return just an Id and Description for a few "lookup" properties of an entity. Removing the () from the Select statement was the key.
1

In some circumstances there simple workaround might be possible: make the type appear as different types. E.g. make 2 subclasses from the original class. This workaround is of course quite dirty but the Linq requirement is artificial by itself. In my case this helped.

Improve this answer

Comments

0

You can put the conditional statement in each property initializer.

var eventData = dbContext.Event.Select(t => new { Address = new AnonymousEventGetAddress { AddressLine1 = includeAddress ? t.Address.AddressLine1 : null, CityName = includeAddress ? t.Address.AddressCityName : null } });
Improve this answer

1 Comment

In case of a list of referenced entities, this will return me an IEnumerable with empty objects instead of an empty IEnumerable. And I still don't get the point why with true it is working, but with includeAddress it's not?
0

In my opinion, I always try to steer away from putting anything more complicated then selecting data into IQueryable because they're Expression, which means they're never executed - they're compiled.

So, I would tackle this problem like so (this has a nice air of simplicity around it):

Create a DTO for the return data:

public class EventDto { // some properties here that you need public Address Address {get;set;} }

Then I would split your logic around the includeAddress

public IEnumerable<EventDto> IncludeAddress(DbContext dbContext) { return dbContext.Event.Select(t => new { // select out your other properties here Address = new { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName }, }).ToList().Select(x => new EventDto { Address = Address }); // put what ever mapping logic you have up there, whether you use AutoMapper or hand map it doesn't matter. }

The method NoAddress or whatever you want to call it will look similar but with no Address, and you map it back.

You can then choose which one simply:

var eventDtos = new List<EventDto>(); if (includeAddress) eventDtos.AddRange(this.IncludeAddress(dbContext)); else eventDtos.AddRange(this.NoAddress(dbContext)); eventDtos.ForEach(e => { if (e.Address == null) e.Address = new Address(); });

If you Select does have a lot of logic in it, then I would consider moving it out into a sproc where it will be easier to read the SQL.

Obviously this is just a guide, gives you an idea on how to tackle the problem.

Improve this answer

Comments

-3

I had the same problem, and found the solution to be, to add .ToList() , before the select-function :

var eventData = dbContext.Event.ToList().Select(t => new { Address = includeAddress ? new AnonymousEventGetAddress { AddressLine1 = t.Address.AddressLine1, CityName = t.Address.AddressCityName } : new AnonymousEventGetAddress(), });
Improve this answer

1 Comment

This causes all the table data is read from the table (not good if you have more than a couple of dozens of rows)

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.