I\'m trying to inner join a users table to itself using an eloquent model. I\'ve looked everywhere but can\'t seem to find a solution to this withou
I came across the same problem quite some time ago and have thus been following this problem closely and have made a lot of research. I have come across some of the solutions you have also found, and some more, and also have thought of other solutions that I summed here, mostly how to get both user_ids in the same column. I am afraid they will all not work well. I am also afraid that using any custom classes will stop you from using all of Laravel's handy relation features (especially eager loading). So I still thought what one could do, and, until one comes up with a hasMany-function on many columns, I think I have come up with a possible solution yesterday. I will show it first and then apply it to your project.
In my project, one user partners with another one (= partnership) and then later will be assigned a commission. So I had the following tables:
USERS
id | name
---------|------
1 | foo
2 | bar
17 | baz
20 | Joe
48 | Jane
51 | Jim
PARTNERSHIPS
id | partner1 | partner2 | confirmed | other_columns
----|-----------|-----------|-----------|---------------
1 | 1 | 2 | 1 |
9 | 17 | 20 | 1 |
23 | 48 | 51 | 1 |
As each user should always have only one active partnership, the non-active being soft-deleted, I could have helped myself by just using the hasMany function twice:
//user.php
public function partnerships()
{
$r = $this->hasMany(Partnership::class, 'partner1');
if(! $r->count() ){
$r = $this->hasMany(Partnership::class, 'partner2');
}
return $r;
}
But if I had wanted to lookup all partnerships of a user, current and past, this of course, wouldn't have worked.
Yesterday, I came up with the solution, that is close to yours, of using a pivot table but with a little difference of using another table:
USERS
(same as above)
PARTNERSHIP_USER
user_id | partnership_id
--------|----------------
1 | 1
2 | 1
17 | 9
20 | 9
48 | 23
51 | 23
PARTNERSHIPS
id | confirmed | other_columns
----|-----------|---------------
1 | 1 |
9 | 1 |
23 | 1 |
// user.php
public function partnerships(){
return $this->belongsToMany(Partnership::class);
}
public function getPartners(){
return $this->partnerships()->with(['users' => function ($query){
$query->where('user_id', '<>', $this->id);
}])->get();
}
public function getCurrentPartner(){
return $this->partnerships()->latest()->with(['users' => function ($query){
$query->where('user_id', '<>', $this->id);
}])->get();
}
// partnership.php
public function users(){
return $this->belongsToMany(User::class);
}
Of course, this comes with the drawback that you always have to create and maintain two entrances in the pivot table but I think this occasional extra load for the database -- how often will this be altered anyway? -- is preferable to having two select queries on two columns every time (and from your example it seemed that you duplicated the entries in your friends table anyway).
In your example the tables could be structured like this:
USERS
id | name
---------|------
1 | foo
2 | bar
3 | baz
FRIENDSHIP_USER
user_id | friendship_id
---------|------
1 | 1
2 | 1
3 | 2
1 | 2
FRIENDSHIPS
id |send_id* | receive_id* | is_blocked | [all the other nice stuff
--------|---------|-------------|------------|- you want to save]
1 | 1 | 2 | 0 |
2 | 3 | 1 | 0 |
[*send_id and receive_id are optional except
you really want to save who did what]
Edit: My $user->partners() looks like this:
// user.php
// PARTNERSHIPS
public function partnerships(){
// 'failed' is a custom fields in the pivot table, like the 'is_blocked' in your example
return $this->belongsToMany(Partnership::class)
->withPivot('failed');
}
// PARTNERS
public function partners(){
// this query goes forth to partnerships and then back to users.
// The subquery excludes the id of the querying user when going back
// (when I ask for "partners", I want only the second person to be returned)
return $this->partnerships()
->with(['users' => function ($query){
$query->where('user_id', '<>', $this->id);
}]);
}