ChatGPT解决这个技术问题 Extra ChatGPT

Entity Framework Code First - two Foreign Keys from same table

I've just started using EF code first, so I'm a total beginner in this topic.

I wanted to create relations between Teams and Matches:

1 match = 2 teams (home, guest) and result.

I thought it's easy to create such a model, so I started coding:

public class Team
{
    [Key]
    public int TeamId { get; set;} 
    public string Name { get; set; }

    public virtual ICollection<Match> Matches { get; set; }
}


public class Match
{
    [Key]
    public int MatchId { get; set; }

    [ForeignKey("HomeTeam"), Column(Order = 0)]
    public int HomeTeamId { get; set; }
    [ForeignKey("GuestTeam"), Column(Order = 1)]
    public int GuestTeamId { get; set; }

    public float HomePoints { get; set; }
    public float GuestPoints { get; set; }
    public DateTime Date { get; set; }

    public virtual Team HomeTeam { get; set; }
    public virtual Team GuestTeam { get; set; }
}

And I get an exception:

The referential relationship will result in a cyclical reference that is not allowed. [ Constraint name = Match_GuestTeam ]

How can I create such a model, with 2 foreign keys to the same table?


L
Ladislav Mrnka

Try this:

public class Team
{
    public int TeamId { get; set;} 
    public string Name { get; set; }

    public virtual ICollection<Match> HomeMatches { get; set; }
    public virtual ICollection<Match> AwayMatches { get; set; }
}

public class Match
{
    public int MatchId { get; set; }

    public int HomeTeamId { get; set; }
    public int GuestTeamId { get; set; }

    public float HomePoints { get; set; }
    public float GuestPoints { get; set; }
    public DateTime Date { get; set; }

    public virtual Team HomeTeam { get; set; }
    public virtual Team GuestTeam { get; set; }
}


public class Context : DbContext
{
    ...

    protected override void OnModelCreating(DbModelBuilder modelBuilder)
    {
        modelBuilder.Entity<Match>()
                    .HasRequired(m => m.HomeTeam)
                    .WithMany(t => t.HomeMatches)
                    .HasForeignKey(m => m.HomeTeamId)
                    .WillCascadeOnDelete(false);

        modelBuilder.Entity<Match>()
                    .HasRequired(m => m.GuestTeam)
                    .WithMany(t => t.AwayMatches)
                    .HasForeignKey(m => m.GuestTeamId)
                    .WillCascadeOnDelete(false);
    }
}

Primary keys are mapped by default convention. Team must have two collection of matches. You can't have single collection referenced by two FKs. Match is mapped without cascading delete because it doesn't work in these self referencing many-to-many.


What if two teams are allowed to play only once?
@NickW: That is something you have to handle in your application and not in the mapping. From the mapping perspective, pairs are allowed to play twice (each is guest and home once).
I have a similar model. What is the proper way to handle cascade delete if a team is removed? I looked into creating an INSTEAD OF DELETE trigger but not sure if there is a better solution? I would prefer to handle this in the DB, not the application.
@mrshickadance: It is the same. One approach uses fluent API and another data annotations.
If I use WillCascadeOnDelete false then If I want to delete the Team Then it is throwing error. A relationship from the 'Team_HomeMatches' AssociationSet is in the 'Deleted' state. Given multiplicity constraints, a corresponding 'Team_HomeMatches_Target' must also in the 'Deleted' state.
S
ShaneA

It's also possible to specify the ForeignKey() attribute on the navigation property:

[ForeignKey("HomeTeamID")]
public virtual Team HomeTeam { get; set; }
[ForeignKey("GuestTeamID")]
public virtual Team GuestTeam { get; set; }

That way you don't need to add any code to the OnModelCreate method


I get the same exception either way.
This is my standard way of specifying foreign keys which works for all cases EXCEPT when an entity contains more than one nav property of the same type (similar to the HomeTeam and GuestTeam scenario), in which case EF gets confused in generating the SQL. Solution is to add code to OnModelCreate as per the accepted answer as well as the two collections for both sides of the relationship.
i use onmodelcreating in all cases except the mentioned case , i use the data annotation foreign key , as well i don't know why it is not accepted !!
k
khoa_chung_89

I know it's a several years old post and you may solve your problem with above solution. However, i just want to suggest using InverseProperty for someone who still need. At least you don't need to change anything in OnModelCreating.

The below code is un-tested.

public class Team
{
    [Key]
    public int TeamId { get; set;} 
    public string Name { get; set; }

    [InverseProperty("HomeTeam")]
    public virtual ICollection<Match> HomeMatches { get; set; }

    [InverseProperty("GuestTeam")]
    public virtual ICollection<Match> GuestMatches { get; set; }
}


public class Match
{
    [Key]
    public int MatchId { get; set; }

    public float HomePoints { get; set; }
    public float GuestPoints { get; set; }
    public DateTime Date { get; set; }

    public virtual Team HomeTeam { get; set; }
    public virtual Team GuestTeam { get; set; }
}

You can read more about InverseProperty on MSDN: https://msdn.microsoft.com/en-us/data/jj591583?f=255&MSPPError=-2147217396#Relationships


Thanks for this answer, however it makes the Foreign Key columns nullable in the Match table.
This worked great for me in EF 6 where nullable collections were needed.
If you want to avoid fluent api (for whatever reason #differentdiscussion) this works fantastically. In my case I needed to include an additional foriegnKey annotation on the "Match" entity, because my fields/tables have strings for PK's.
This worked greatly for me. Btw. if you do not want the columns nullable you can just specify foreign key with the [ForeignKey] attribute. If the key is not nullable then you are all set.
M
Maico

You can try this too:

public class Match
{
    [Key]
    public int MatchId { get; set; }

    [ForeignKey("HomeTeam"), Column(Order = 0)]
    public int? HomeTeamId { get; set; }
    [ForeignKey("GuestTeam"), Column(Order = 1)]
    public int? GuestTeamId { get; set; }

    public float HomePoints { get; set; }
    public float GuestPoints { get; set; }
    public DateTime Date { get; set; }

    public virtual Team HomeTeam { get; set; }
    public virtual Team GuestTeam { get; set; }
}

When you make a FK column allow NULLS, you are breaking the cycle. Or we are just cheating the EF schema generator.

In my case, this simple modification solve the problem.


Caution readers. Although this might work-around the schema definition problem, it alters the semantics. It is probably not the case that a Match can be had without two teams.
Y
Yossarian

This is because Cascade Deletes are enabled by default. The problem is that when you call a delete on the entity, it will delete each of the f-key referenced entities as well. You should not make 'required' values nullable to fix this problem. A better option would be to remove EF Code First's Cascade delete convention:

modelBuilder.Conventions.Remove<OneToManyCascadeDeleteConvention>(); 

It's probably safer to explicitly indicate when to do a cascade delete for each of the children when mapping/config. the entity.


So what is it after this is executed? Restrict instead of Cascade?
p
pritesh agrawal

InverseProperty in EF Core makes the solution easy and clean.

InverseProperty

So the desired solution would be:

public class Team
{
    [Key]
    public int TeamId { get; set;} 
    public string Name { get; set; }

    [InverseProperty(nameof(Match.HomeTeam))]
    public ICollection<Match> HomeMatches{ get; set; }

    [InverseProperty(nameof(Match.GuestTeam))]
    public ICollection<Match> AwayMatches{ get; set; }
}


public class Match
{
    [Key]
    public int MatchId { get; set; }

    [ForeignKey(nameof(HomeTeam)), Column(Order = 0)]
    public int HomeTeamId { get; set; }
    [ForeignKey(nameof(GuestTeam)), Column(Order = 1)]
    public int GuestTeamId { get; set; }

    public float HomePoints { get; set; }
    public float GuestPoints { get; set; }
    public DateTime Date { get; set; }

    public Team HomeTeam { get; set; }
    public Team GuestTeam { get; set; }
}

Why do we need 'Column(Order = 0)'? This solved my issue by the way
It is used to order columns. Copied from the original question. It has nothing to do with Inverse Property.
U
U.Y.

I know this is pretty old question but coming here in 2021 with EF Core > 3 solution below worked for me.

Make sure to make foreign keys nullable Specify default behavior on Delete public class Match { public int? HomeTeamId { get; set; } public int? GuestTeamId { get; set; } public float HomePoints { get; set; } public float GuestPoints { get; set; } public DateTime Date { get; set; } public Team HomeTeam { get; set; } public Team GuestTeam { get; set; } } protected override void OnModelCreating(ModelBuilder modelBuilder) { modelBuilder.Entity() .HasRequired(m => m.HomeTeam) .WithMany(t => t.HomeMatches) .HasForeignKey(m => m.HomeTeamId) .OnDelete(DeleteBehavior.ClientSetNull); modelBuilder.Entity() .HasRequired(m => m.GuestTeam) .WithMany(t => t.AwayMatches) .HasForeignKey(m => m.GuestTeamId) .OnDelete(DeleteBehavior.ClientSetNull); }