Editing this question in the hope to make it clearer.
We have entity framework code first setup. I've simplified two classes for the purposes of example, in reality there are around 10+ more classes similar to the 'Record', where Item is a navigational property/foreign key.
Item class:
public class Item
{
public int Id { get; set; }
public int AccountId { get; set; }
public List<UserItemMapping> UserItemMappings { get; set; }
public List<GroupItemMapping> GroupItemMappings { get; set; }
}
Record class:
public class Record
{
public int ItemId { get; set; }
public Item Item { get; set; }
}
this.User is an injected user object into each repo and is contained on the repository base.
We have an Item repository with the following code:
var items = this.GetAll()
.Where(i => i.AccountId == this.User.AccountId);
I created the follow expression on the repository base to easily filter on that (in the hope of re-use). We cannot use static extension methods due to how LINQ to entities works (System.NotSupportedException "LINQ to Entities does not recognize the method X and this method cannot be translated into a store expression.").
protected Expression<Func<Item, bool>> ItemIsOnAccount()
{
return item => item.AccountId == this.User.AccountId;
}
I have solved the case of the above, by doing this:
var items = this.GetAll().Where(this.ItemIsOnAccount());
We have additional filtering based on user permissions within that account (again, another case where I do not want to repeat this code in every repo we have):
protected Expression<Func<Item, bool>> SubUserCanAccessItem()
{
return item => this.User.AllowAllItems
|| item.UserItemMappings.Any(d => d.UserId.Value == this.User.Id)
|| item.GroupItemMappings.Any(vm =>
vm.Group.GroupUserMappings
.Any(um => um.UserId == this.User.Id));
}
Which I am able to use as follows:
var items = this.GetAll().Where(this.SubUserCanAccessItem());
However, what we also need, in the Record repository is a way to solve the following:
var records = this.GetAll()
.Where(i => i.Item.AccountId == this.User.AccountId);
Because Item is a single navigational property, I do not know how to apply the expressions I have created to this object.
I want to reuse the expression I created in the repo base on all of these other repos, so that my 'permission based' code is all in the same place, but I cannot simply throw it in because the Where clause in this case is of Expression< Func < Record,bool >>.
Creating an interface with a method of:
Item GetItem();
on it and putting it on the Record class does not work because of LINQ to entities.
I cannot also create a base abstract class and inherit from it, because there could be other objects than Item that need to be filtered on. For instance a Record could also have a 'Thing' on it that has permission logic. Not all objects will require to be filtered by 'Item' and 'Thing', some by only one, some by another, some by both:
var items = this.GetAll()
.Where(this.ItemIsOnAccount())
.Where(this.ThingIsOnAccount());
var itemType2s = this.GetAll().Where(this.ThingIsOnAccount());
var itemType3s = this.GetAll().Where(this.ItemIsOnAccount());
Due to this having a single parent class would not work.
Is there a way in which I can reuse the expressions I have already created, or at least create an expression/modify the originals to work across the board within the OTHER repos that of course return their own objects in a GetAll, but all have a navigation property to Item? How would I need to modify the other repos to work with these?
Thanks
The first step for expression reusability is to move the expressions to a common static class. Since in your case they are tied to User, I would make them User extension methods (but note that they will return expressions):
public static partial class UserFilters
{
public static Expression<Func<Item, bool>> OwnsItem(this User user)
=> item => item.AccountId == user.AccountId;
public static Expression<Func<Item, bool>> CanAccessItem(this User user)
{
if (user.AllowAllItems) return item => true;
return item => item.UserItemMappings.Any(d => d.UserId.Value == user.Id) ||
item.GroupItemMappings.Any(vm => vm.Group.GroupUserMappings.Any(um => um.UserId == user.Id));
}
}
Now the Item repository would use
var items = this.GetAll().Where(this.User.OwnsItem());
or
var items = this.GetAll().Where(this.User.CanAccessItem());
In order to be reusable for entities having Item reference, you would need a small helper utility for composing lambda expressions from other lambda expressions, similar to Convert Linq expression "obj => obj.Prop" into "parent => parent.obj.Prop".
It's possible to implement it with Expression.Invoke, but since not all query providers support for invocation expressions (EF6 doesn't for sure, EF Core does), as usual we'll use a custom expression visitor for replacing a lambda parameter expression with another arbitrary expression:
public static partial class ExpressionUtils
{
public static Expression ReplaceParameter(this Expression expression, ParameterExpression source, Expression target)
=> new ParameterReplacer { Source = source, Target = target }.Visit(expression);
class ParameterReplacer : ExpressionVisitor
{
public ParameterExpression Source;
public Expression Target;
protected override Expression VisitParameter(ParameterExpression node)
=> node == Source ? Target : node;
}
}
And the two composing functions are as follows (I don't like the name Compose, so sometimes I use the name Map, sometimes Select, Bind, Transform etc., but functionally they do the same. In this case I'm using Apply and ApplyTo, with the only difference being the transformation direction):
public static partial class ExpressionUtils
{
public static Expression<Func<TOuter, TResult>> Apply<TOuter, TInner, TResult>(this Expression<Func<TOuter, TInner>> outer, Expression<Func<TInner, TResult>> inner)
=> Expression.Lambda<Func<TOuter, TResult>>(inner.Body.ReplaceParameter(inner.Parameters[0], outer.Body), outer.Parameters);
public static Expression<Func<TOuter, TResult>> ApplyTo<TOuter, TInner, TResult>(this Expression<Func<TInner, TResult>> inner, Expression<Func<TOuter, TInner>> outer)
=> outer.Apply(inner);
}
(Nothing special there, code provided for completeness)
Now you could reuse the original filters by "applying" them to a expression which selects Item property from another entity:
public static partial class UserFilters
{
public static Expression<Func<T, bool>> Owns<T>(this User user, Expression<Func<T, Item>> item)
=> user.OwnsItem().ApplyTo(item);
public static Expression<Func<T, bool>> CanAccess<T>(this User user, Expression<Func<T, Item>> item)
=> user.CanAccessItem().ApplyTo(item);
}
and add the following to the entity repository (in this case, Record repository):
static Expression<Func<Record, Item>> RecordItem => entity => entity.Item;
which would allow you to use there
var records = this.GetAll().Where(this.User.Owns(RecordItem));
or
var records = this.GetAll().Where(this.User.CanAccess(RecordItem));
This should be enough to satisfy your requirements.
You can go further and define an interface like this
public interface IHasItem
{
Item Item { get; set; }
}
and let the entities implement it
public class Record : IHasItem // <--
{
// Same as in the example - IHasItem.Item is auto implemented
// ...
}
then add additional helpers like this
public static partial class UserFilters
{
public static Expression<Func<T, Item>> GetItem<T>() where T : class, IHasItem
=> entity => entity.Item;
public static Expression<Func<T, bool>> OwnsItem<T>(this User user) where T : class, IHasItem
=> user.Owns(GetItem<T>());
public static Expression<Func<T, bool>> CanAccessItem<T>(this User user) where T : class, IHasItem
=> user.CanAccess(GetItem<T>());
}
which would allow you omit the RecordItem expression in the repository and use this instead
var records = this.GetAll().Where(this.User.OwnsItem<Record>());
or
var records = this.GetAll().Where(this.User.CanAccessItem<Record>());
Not sure if it gives you a better readability, but is an option, and syntactically is closer to Item methods.
For Thing etc. just add similar UserFilters methods.
As a bonus, you can go even further and add the usual PredicateBuilder methods And and Or
public static partial class ExpressionUtils
{
public static Expression<Func<T, bool>> And<T>(this Expression<Func<T, bool>> left, Expression<Func<T, bool>> right)
=> Expression.Lambda<Func<T, bool>>(Expression.AndAlso(left.Body,
right.Body.ReplaceParameter(right.Parameters[0], left.Parameters[0])), left.Parameters);
public static Expression<Func<T, bool>> Or<T>(this Expression<Func<T, bool>> left, Expression<Func<T, bool>> right)
=> Expression.Lambda<Func<T, bool>>(Expression.OrElse(left.Body,
right.Body.ReplaceParameter(right.Parameters[0], left.Parameters[0])), left.Parameters);
}
so you could use something like this if needed
var items = this.GetAll().Where(this.User.OwnsItem().Or(this.User.CanAccessItem()));
in the Item repository, or
var records = this.GetAll().Where(this.User.OwnsItem<Record>().Or(this.User.CanAccessItem<Record>()));
in the Record repository.
I can't really tell if this could work in your case, depends on how your entities might be setup, but one thing you can try is to have an interface like IHasItemProperty with a GetItem() method and have the entities where you want to use this implement that interface. Something like this :
public interface IHasItemProperty {
Item GetItem();
}
public class Item: IHasItemProperty {
public Item GetItem() {
return this;
}
public int UserId {get; set;}
}
public class Record: IHasItemProperty {
public Item item{get;set;}
public Item GetItem() {
return this.item;
}
}
public class Repo
{
protected Expression<Func<T, bool>> ItemIsOnAccount<T>() where T: IHasItemProperty
{
return entity => entity.GetItem().UserId == 5;
}
}
I have used an int just to make things simpler.
You should be able to do this with .AsQueryable().
class Account
{
public IEnumerable<User> Users { get; set; }
public User SingleUser { get; set; }
static void Query()
{
IQueryable<Account> accounts = new Account[0].AsQueryable();
Expression<Func<User, bool>> userExpression = x => x.Selected;
Expression<Func<Account, bool>> accountAndUsersExpression =
x => x.Users.AsQueryable().Where(userExpression).Any();
var resultWithUsers = accounts.Where(accountAndUsersExpression);
Expression<Func<Account, bool>> accountAndSingleUserExpression =
x => new[] { x.SingleUser }.AsQueryable().Where(userExpression).Any();
var resultWithSingleUser = accounts.Where(accountAndSingleUserExpression);
}
}
class User
{
public bool Selected { get; set; }
}
You should only use sql (or your database like) items for the predicate. If you put this.User.AccountId into your lambda, that does not exists at database and can't be parsed by it, that's the source of your error message.
Related
I'm using EntityFrameworkCore and am trying to create a simplified instance of searching for either 'equal to' or 'like' based whether the search object contians the wildcard character.
Here's the base of what I'm working with
public class Person
{
public string Name;
public string MothersName;
public string FathersName;
}
public class SearchPerson
{
public string Name;
}
public class Program
{
public void FindPerson(SearchPerson searchPerson)
{
if (!string.IsNullOrEmpty(searchPerson.Name))
{
if (searchPerson.Name.Contains("%"))
{
EFPersonObject.Where(m => EF.Functions.Like(m.Name, searchPerson.Name));
}
else
{
EFPersonObject.Where(m => m.Name == searchPerson.Name);
}
}
}
}
If my SearchPerson class extends to 5 or 10 or 15 possible search params, there is a lot of repeated code. I should be able to implement some reflection in an extension and using Jim C's response here, get and pass the name of the property and simplify a lot of it down to one line
public static class SearchExtension
{
public static void FindLike<T>(this DbSet<T> model, PropertyInfo info, string searchValue) where T : class
{
if (!string.IsNullOrEmpty(searchValue))
{
if (searchValue.Contains("%"))
{
model.Where(m => EF.Functions.Like(typeof(T).GetProperty(info.Name).GetValue(model, null).ToString(), searchValue));
}
else
{
model.Where(m => typeof(T).GetProperty(info.Name).GetValue(model, null).ToString() == searchValue);
}
}
}
}
Usage:
EFPersonObject.FindLike(typeof(Person).GetProperty(RemoteMgr.GetPropertyName(()=>typeof(Person).Name)), searchPerson.Name);
(I haven't tested it yet, but if it isn't right, it should be close), but I'm going to assume I'm going to take a performance hit. Is there another way to implement this where reflection isn't needed to avoid the performance hit?
Using reflection (and other non SQL translatable) calls inside the query expression tree is not a good idea. In EF Core 1x and 2.x it will cause client evaluation, and EF Core v3+ will throw exception similar to EF 6.
LINQ to Entities best work with expressions. And once you need expression, you'd better make your custom extension method receive lambda expression directly rather than PropertyInfo obtained via lambda expression as in the linked topic.
Here is a sample implementation of the above:
public static partial class QueryableExtensions
{
public static IQueryable<T> WhereMatch<T>(this IQueryable<T> source, Expression<Func<T, string>> expr, string searchValue)
{
if (string.IsNullOrEmpty(searchValue))
return source;
else if (searchValue.Contains("%"))
return source.Where(expr.Map(value => EF.Functions.Like(value, searchValue)));
else
return source.Where(expr.Map(value => value == searchValue));
}
static Expression<Func<TSource, TTarget>> Map<TSource, TIntermediate, TTarget>(this Expression<Func<TSource, TIntermediate>> source, Expression<Func<TIntermediate, TTarget>> target)
=> Expression.Lambda<Func<TSource, TTarget>>(Expression.Invoke(target, source.Body), source.Parameters);
}
The main method is WhereMatch. It uses a small Expression helper method called Map for composing lambda expressions from other lambda expressions.
Sample usage would be:
// SearchPerson searchPerson
// DbContext db
var query = db.Set<Person>()
.WhereMatch(p => p.Name, searchPerson.Name)
.WhereMatch(p => p.MothersName, searchPerson.MothersName)
.WhereMatch(p => p.FathersName, searchPerson.FathersName);
For Equality comparison you should use ==:
EFPersonObject.Where(m => m.Name == searchPerson.Name);
For LIKE :
like 'something%': (StartsWith Method)
EFPersonObject.Where(m => m.Name.StartsWith(searchPerson.Name));
like '%something': (EndsWith Method)
EFPersonObject.Where(m => m.Name.EndsWith(searchPerson.Name));
like '%something%': (Contains Method)
EFPersonObject.Where(m => m.Name.Contains(searchPerson.Name));
I want to add a Guid identifier to a few of my entities.
My initial approach was to make it the Primary Key for those entities. But that seems to make the rest of my code more complex.
For example we must now allow for multiple ID types in the generic repository and specify the PK type for each DbEntity (code samples below).
Therefore I am considering maintaining a consistent int PK on all entities, and adding the Guid as an additional property when required. This also means that I don't have to worry about ensuring the Guid is not used as a clustering key.
Simpler code. No cluster worries. The 'additional property' approach looks like a winner to me.
However I don't see it mentioned, used or recommended anywhere. (In comparison there are plenty of articles disussing using a Guid for a PK).
Are there any disadvantages to this approach? Could doing things this way come back and bite me?
Using Only Int as PK
public class DbEntity
{
[Key]
public int ID { get; set; }
}
public class Car : DbEntity
{}
public class House: DbEntity
{}
public virtual T SelectByID(int id, params Expression<Func<T, object>>[] includeExpressions)
{
var set = includeExpressions.Aggregate<Expression<Func<T, object>>, IQueryable<T>>
(table, (current, expression) => current.Include(expression));
return set.SingleOrDefault(i => i.ID == id);
}
public virtual T SelectByGuid(string guid, params Expression<Func<T, object>>[] includeExpressions)
{
var set = includeExpressions.Aggregate<Expression<Func<T, object>>, IQueryable<T>>
(table, (current, expression) => current.Include(expression));
return set.SingleOrDefault(i => i.AdditionalGuidProperty == guid);
}
Using Both Int and Guid as PK
public class DbEntity<PKT>
{
[Key]
public PKT ID { get; set; }
}
public class Car : DbEntity<int>
{}
public class House : DbEntity<guid>
{}
public virtual T SelectByID(PKT id, params Expression<Func<T, object>>[] includeExpressions)
{
var set = includeExpressions.Aggregate<Expression<Func<T, object>>, IQueryable<T>>
(table, (current, expression) => current.Include(expression));
ParameterExpression parameter = Expression.Parameter(typeof(T), "s");
PropertyInfo propertyInfo = typeof(T).GetProperty("ID");
MemberExpression memberExpression = Expression.MakeMemberAccess(parameter, propertyInfo);
ConstantExpression constantExpression = Expression.Constant(id, typeof(PKT));
BinaryExpression binaryExpression = Expression.Equal(memberExpression, constantExpression);
Expression<Func<T, bool>> lambda = Expression.Lambda<Func<T, bool>>(binaryExpression, parameter);
return set.SingleOrDefault(lambda);
}
I have a data access class that acts as an intermediary between logic classes and the underlying datasource, which is interchangeable. This class allows you to query the datasource using lambdas, LINQ-style. A source-agnostic class provides high-level functionality powered by a few basic operations (Add, GetAll, Update, Delete, Commit) that are implemented by small adapter classes, one for each source type (SQL, SQlite, XML serialiser, WCF client, REST client, whatever).
My problem is that some relational data sources (particularly SQLite) aren't smart enough to load relationship properties when I need them; I have to explicitly ask for them to be included. This is fine for my Get methods; I can pass a params array of expressions to load anything I need. With .Any(), however, this feels a bit odd - if I'm asking if there are any Customer records whose Purchases list contains a certain item, I shouldn't then have to tell it to load the Purchases list; that seems like the sort of thing it should be able to figure out.
So my Any() method takes Expression<Func<T, bool>> where T is obviously going to be the type I'm operating on. In the above example, it'd be used something like this:
using (var db = _dataAccessProvider.NewTransaction())
{
return db.Any<Customer>(c => c.Purchases.Contains(someProduct));
}
Is it possible to take the Expression<Func<Customer, bool>> that represents the operation c => c.Purchases.Contains(someProduct)) and work out that the property it's referring to is c => c.Purchases? How would I go about doing that? What about a lambda that touches multiple properties?
Use ExpressionVisitor to find all MemberExpression expressions which reference required object properties.
Quick example:
using System;
using System.Collections.Generic;
using System.Linq.Expressions;
using System.Reflection;
class Program
{
sealed class ReferencedPropertyFinder : ExpressionVisitor
{
private readonly Type _ownerType;
private readonly List<PropertyInfo> _properties = new List<PropertyInfo>();
public ReferencedPropertyFinder(Type ownerType)
{
_ownerType = ownerType;
}
public IReadOnlyList<PropertyInfo> Properties
{
get { return _properties; }
}
protected override Expression VisitMember(MemberExpression node)
{
var propertyInfo = node.Member as PropertyInfo;
if(propertyInfo != null && _ownerType.IsAssignableFrom(propertyInfo.DeclaringType))
{
// probably more filtering required
_properties.Add(propertyInfo);
}
return base.VisitMember(node);
}
}
private static IReadOnlyList<PropertyInfo> GetReferencedProperties<T, U>(Expression<Func<T, U>> expression)
{
var v = new ReferencedPropertyFinder(typeof(T));
v.Visit(expression);
return v.Properties;
}
sealed class TestEntity
{
public int PropertyA { get; set; }
public int PropertyB { get; set; }
public int PropertyC { get; set; }
}
static void Main(string[] args)
{
Expression<Func<TestEntity, int>> expression =
e => e.PropertyA + e.PropertyB;
foreach(var property in GetReferencedProperties(expression))
{
Console.WriteLine(property.Name);
}
}
}
My question is very similar to the following two questions but I have an added requirement that these do not satisfy.
How to set property value using Expressions?
How set value a property selector Expression<Func<T,TResult>>
Just like those questions, I have an Expression<Func<TEntity, TProperty>> where I want to set a value to the specified property. And those solutions work great if the body of the expression is only one level deep, such as x => x.FirstName but they don't work at all if that body is deeper, like x => x.Parent.FirstName.
Is there some way to take this deeper expression and set the value to it? I don't need a terribly robust execution-deferred solution but I do need something that I can execute on an object and it would work whether 1-level or multiple levels deep. I also need to support most typical types you'd expect in a database (long, int?, string, Decimal, DateTime?, etc. although I don't care about more complex things like geo types).
For conversation's sake, let's say we're working with these objects although assume we need to handle N levels deep, not just 1 or 2:
public class Parent
{
public string FirstName { get; set; }
}
public class Child
{
public Child()
{
Mom = new Parent(); // so we don't have to worry about nulls
}
public string FavoriteToy { get; set; }
public Parent Mom { get; set; }
}
and let's say this is our unit test:
[TestFixture]
public class Tests
{
[Test]
public void MyTest()
{
var kid = new Child();
Expression<Func<Child, string>> momNameSelector = (ch => ch.Mom.FirstName);
Expression<Func<Child, string>> toyNameSelector = (ch => ch.FavoriteToy);
kid.ExecuteMagicSetter(momNameSelector, "Jane");
kid.ExecuteMagicSetter(toyNameSelector, "Bopp-It!");
Assert.That(kid.Mom.FirstName, Is.EqualTo("Jane"));
Assert.That(kid.FavoriteToy, Is.EqualTo("Bopp-It!"));
}
}
and our extension method we're looking at (I'm not set on it needing to be an extension method but it seems simple enough) would look like this:
public static TEntity ExecuteMagicSetter<TEntity, TProperty>(this TEntity obj, Expression<Func<TEntity, TProperty>> selector, TProperty value)
where TEntity : class, new() // I don't require this but I can allow this restriction if it helps
{
// magic
}
P.S. This version of code was written in the SO editor - my apologies for dumb syntax issues but this should be darn close! #LockedDownWorkstationsSuck
As I stated in the comments, it shouldn't be all that complicated. With the selector, just add an assignment to the expression. You'll just need to compile and run the expression.
public static TEntity ExecuteMagicSetter<TEntity, TProperty>(
this TEntity obj,
Expression<Func<TEntity, TProperty>> selector,
TProperty value)
{
var setterExpr = CreateSetter(selector);
setterExpr.Compile()(obj, value);
return obj;
}
private static Expression<Action<TEntity, TProperty>> CreateSetter<TEntity, TProperty>
(Expression<Func<TEntity, TProperty>> selector)
{
var valueParam = Expression.Parameter(typeof(TProperty));
var body = Expression.Assign(selector.Body, valueParam);
return Expression.Lambda<Action<TEntity, TProperty>>(body,
selector.Parameters.Single(),
valueParam);
}
I'm trying to build a generic method that EF4.1 to look in both the Database and the Local memory for a particular row in a table that matches a particular criteria.
So far, this is what I have this.
This is the caller.
dbEntities.MyTables.LocalAndDb(delegate(MyTable s)
{ return s.Description.Contains("test"); });
This is LocalAndDb
public static object LocalAndDb<T>(this DbSet<T> myTable, Func<T, bool> function) where T : class
{
// look in local
var item = myTable.Local.Where(o => function((T)o)).FirstOrDefault()
// if not exist, look in the database
if (item == null)
{
Expression<Func<T, bool>> predicate = (u) => function(u);
item = myTable.Where(predicate).FirstOrDefault();
}
return item;
}
The problem is with this line.
item = myTable.Where(predicate).FirstOrDefault();
When it calls the database, it throws this error.
"The LINQ expression node type 'Invoke' is not supported in LINQ to Entities."
I imagine it's because I'm passing in an anonymous method and it doesn't know how to turn this into SQL. I thought converting it to an Expression object would do the trick but it's still not working for me.
What do I need to do to make a anonymous method become something that LINQ can turn into SQL?
To make this work, you need to pass the lambda expression to LocalAndDb as an expression tree (so that LINQ to Entities can analyze the code and translate it to SQL):
public static object LocalAndDb<T>(this DbSet<T> myTable,
Expression<Func<T, bool>> expr) where T : class {
// ...
if (item == null) {
item = myTable.Where(expr).FirstOrDefault();
}
return item;
}
Then, of course, the problem is that you cannot execute the expression tree when checking the in-memory data. One way to solve this is to use the Compile method of Expression<T>, but that will be a bit inefficient (depending on your scenario).
Another option is to just pass the condition as both function and expression tree:
public static object LocalAndDb<T>(this DbSet<T> myTable,
Func<T, boo> function, Expression<Func<T, bool>> expr) where T : class {
var item = myTable.Local.Where(o => function((T)o)).FirstOrDefault();
if (item == null) {
item = myTable.Where(expr).FirstOrDefault();
}
return item;
}
table.LocalAndDb(t => t.Foo > 10, t => t.Foo > 10);
This is a bit ugly, but it doesn't require inefficient compilation at runtime. If you want a slightly more sophisticated solution, then you can define your own type to keep pre-compiled functions:
class Precompiled<T1, T2> {
public Precompiled(Expression<Func<T1, T2>> expr) {
this.Expression = expr;
this.Function = expr.Compile();
}
public Expression<Func<T1,T2>> Expression { get; private set; }
public Func<T1,T2> Function { get; private set; }
}