[5.8] Allow belongsToMany to take Model/Pivot class name as a second parameter

[linaspasv]

Overview

The motivation behind this RP is to allow a non-hardcoded intermediate table name assignments in belongsToMany relationship. The proposed PR does not alter old behavior but improves and provides more integrity for people using Pivot models.

Example

namespace App\Models;

class Customer extends Model
{
    protected $table = 'customers';
}

class CustomerProfile extends Pivot
{
    protected $table = 'customers_profiles';
}

Example of current usage (hardcoded):

return $this->belongsToMany(Profile::class, 'customers_profiles')->using(CustomerProfile::class);

Example of current usage (dynamic):

$pivotClass = CustomerProfile::class; 
return $this->belongsToMany(Profile::class, (new $pivotClass)->getTable())->using($pivotClass);

Example of (possible) usage after PR (improved behavior does not require to define Pivot through using):

return $this->belongsToMany(Profile::class, CustomerProfile::class);

[mfn]

I don't understand why. The current code is perfectly fine, isn't it? Only to save a few lines?

[linaspasv]

@mfn it's useful for someone who uses Pivot models, especially with custom intermediate table names. As I have noted the old behavior is not changed and there is no breaking change introduced. Pivot class is an extension of a Model class, and until now setting protected $table = '.....' attribute in Pivot models did not have much usage. Finally, the less hardcoded parts we have the more easier is to maintain our applications.

I hope I have cleared your doubts that it's not just to save a few lines (even so Code Less. Do More.)

edit
you could achieve the same behavior with the code below if you would love to approach in a non-hardcoded way but it just does not look right.

$pivotClass = CustomerProfile::class; 
return $this->belongsToMany(Profile::class, (new $pivotClass)->getTable())->using($pivotClass);

[staudenmeir]

What's a scenario where this line (212) is reached?

[linaspasv]

Line (212) is reached when a given table name can't be resolved as an instance of a Model class, so in all other cases it just fallback to the old behavior.

[staudenmeir]

So for the rare case that a table name coincidentally is also a valid class name?

[linaspasv]

My code checks if the given table name is a valid class name but also extends Model / Pivot class. The only rare case I could think of could be when user has a Pivot class named "customers_profiles " without a namespace and uses "customers_profiles" string as a second parameter for belongsToMany:

// note: no namespace used

class customers_profiles extends Pivot
{
    protected $table = 'customers_profiles';
}

......

return $this->belongsToMany(Profile::class, 'customers_profiles')

In this case, my code will see that's the valid Pivot class, will try to load Pivot model and take table name through getTable() method.

dd((new customers_profiles)->getTable()); 
// gives back "customers_profiles" string, 
// or 
// auto-generated value "customers_profile" when user has not defined $table attribute

All other cases I could think of should fallback to an old behavior.

I think the following very rare case should not be an issue. Let me know if I missed something.

Thanks.

[linaspasv]

p.s. I have slightly adjusted the code flow / naming.

[mfn]

you could achieve the same behavior with the code below if you would love to approach in a non-hardcoded way but it just does not look right.

Thanks, makes sense to me 👍

[driesvints]

@linaspasv it would make more chance of being accepted if the PR contains a test as well.