feat: replace proxy mechanism with PHP 8.4 lazy objects

[nikophil]

relates to #899

I think we'll release this in Foundry 2.6

[kbond]

When called without args, let's make the refresh() function proxify everything.

[nikophil]

When called without args, let's make the refresh() function proxify everything.

this feels weird, refresh() function always returns something:

function refresh(object &$object): object
{
    return Configuration::instance()->persistence()->refresh($object);
}

maybe we could add a new function?

[kbond]

this feels weird, refresh() function always returns something:

Hmm yeah, forgot about that. maybe a refresh_all()

[nikophil]

refresh_all() done 👍

[nikophil]

OK CI is now green 🙂

I've deprecated PersistentProxyObjectFactory for PHP 8.4 and handled the deprecations in our testsuite

I've listed in an issue what needs to be done before releasing this

Benchmark is showing a small performance downgrade, which seems totally legit, because we're performing more requests than before with PersistentObjectFactory, but I'm pretty sure that "new Proxy 8.4 Factory" VS the old PersistentProxyObjectFactory (which basically everybody uses) is equivalent in performance, or maybe better

+--------------------------------+-------------------+-----+------+-----+-----------------+------------------+----------------+
| benchmark                      | subject           | set | revs | its | mem_peak        | mode             | rstdev         |
+--------------------------------+-------------------+-----+------+-----+-----------------+------------------+----------------+
| ContactFactoryBench            | bench_create      |     | 10   | 5   | 17.728mb +0.24% | 2.038ms +0.61%   | ±5.23% +32.84% |
| ContactFactoryBench            | bench_create_many | 1   | 10   | 5   | 17.730mb +0.24% | 2.185ms +7.07%   | ±3.36% +45.21% |
| ContactFactoryBench            | bench_create_many | 10  | 10   | 5   | 18.610mb +0.49% | 10.736ms +2.39%  | ±1.80% +32.57% |
| ContactFactoryBench            | bench_create_many | 50  | 10   | 5   | 22.730mb +1.06% | 46.839ms -0.83%  | ±1.03% +1.36%  |
| CategoryFactoryBench           | bench_create      |     | 10   | 5   | 17.945mb +0.33% | 5.206ms +6.12%   | ±1.53% -13.05% |
| CategoryFactoryBench           | bench_create_many | 1   | 10   | 5   | 17.957mb +0.33% | 5.490ms +2.20%   | ±3.03% -15.99% |
| CategoryFactoryBench           | bench_create_many | 10  | 10   | 5   | 21.388mb +1.03% | 38.583ms +0.33%  | ±0.67% -46.50% |
| CategoryFactoryBench           | bench_create_many | 50  | 10   | 5   | 36.158mb +2.20% | 182.886ms -0.06% | ±0.47% -49.97% |
| PersistentDocumentFactoryBench | bench_create      |     | 10   | 5   | 16.240mb +0.24% | 746.168μs -1.00% | ±2.78% +83.50% |
| PersistentDocumentFactoryBench | bench_create_many | 1   | 10   | 5   | 16.245mb +0.24% | 774.627μs -0.07% | ±1.48% -11.57% |
| PersistentDocumentFactoryBench | bench_create_many | 10  | 10   | 5   | 16.310mb +0.74% | 4.824ms -1.42%   | ±4.77% +13.25% |
| PersistentDocumentFactoryBench | bench_create_many | 50  | 10   | 5   | 17.943mb +1.07% | 22.993ms +0.52%  | ±1.01% -45.65% |
+--------------------------------+-------------------+-----+------+-----+-----------------+------------------+----------------+

[kbond]

Looking good! Thanks for working on this!

[nikophil]

@kbond I think this PR is ready - still needs some work to be done (see #899), but in another PR

[kbond]

Looking good! Think we should release 2.6 before merging?

[kbond]

I feel like we should stick with our current snake_case function naming.

In a future PR, thinking we should add some helper phpunit methods to either Factories or a new trait:

• ->save()
• ->assertPersisted()
• etc...

[nikophil]

I feel like we should stick with our current snake_case function naming.

do you mean I should change to _assertPersisted()?

In a future PR, thinking we should add some helper phpunit methods to either Factories or a new trait

yeah, why not

[kbond]

do you mean I should change to _assertPersisted()?

assert_persisted() to match the other functions like persistent_factory()

[nikophil]

LOL ok sorry 😅 🤦