Lazy-load on a per-property base (removes doctrine proxies, replaced by ProxyManager)

[Ocramius]

This is just a proof of concept that I hacked together real quick.

This changes how proxies are generated and lazy-loaded in a very radical way:

• method calls do not cause lazy-loading
• lazy-loading is triggered by access to non-transient properties (things that need to be lazy-loaded)
• access to identifiers or non-mapped properties won't cause lazy-loading
• you can now use $class = get_class($proxy); $object = new $class(); without any side-effects, as the constructor logic is preserved

Following BC breaks need to be fixed or discussed before going forward on this idea:

• proxies don't implement Doctrine\Common\Proxy\Proxy anymore, but ProxyManager\Proxy\GhostObjectInterface (BC break, needs fixing, can be easily done with some effort) BC break is voluntary
• serialize($proxy) now causes proxy initialization (probably needs fixing, as this is a major BC break) - BC break is voluntary

Pending TODOs:

• cloning a proxy is still not fully supported (requires dedicated logic in __clone)
• this is just a PoC, so the code that writes proxies to disk is not yet in place
• hhvm compatibility (HHVM compatibility Ocramius/ProxyManager#210, Closures do not report their class in stack traces facebook/hhvm#4560)
• ProxyManager 2.0.0 is required first (release highly depends on this issue)
• Lots of dead code to remove

To give you an example, it is impossible (without reflection) to cause lazy-loading for the following class:

class Counter
{
    private $id;
    private $otherUnrelatedThing;
    public function doThings() { /* empty body (on purpose) */ }
}

[doctrinebot]

Hello,

thank you for creating this pull request. I have automatically opened an issue
on our Jira Bug Tracker for you. See the issue link:

http://www.doctrine-project.org/jira/browse/DDC-3481

We use Jira to track the state of pull requests and the versions they got
included in.

[goetas]

hi! If you are considering to release a 3.0 version:

• please consider also [3.0] Single entity flush #1139 (that breaks compatibility but it makes sense in my opinion)
• if this will be merged, Always allow proxies on ToOne associations #1141 and Short gettter should work with associations common#338 will not be required anymore

[Ocramius]

I'm not yet considering 3.0. This is just a PR that I opened that is obviously incompatible with 2.x for now.

3.0 will happen, but only after 2.5 and further discussion about what has to land in it.

[stof]

serialize($proxy) now causes proxy initialization (probably needs fixing, as this is a major BC break)

IMO, this is fine. If your entity already implements Serializable, it is already initialized when serializing (because of a method call). And if it does not and the proxy is not initialized yet, unserializing it gives you a broken object (because you don't have the initializer closure to initialize the proxy)

[stof]

However, the Proxy interface BC break is an issue. Couldn't we still implement this interface ?

[Ocramius]

@stof my idea is to modify the code generators to align to the existing interface: not yet sure how.

As for serialize(), the issue is a bit more serious right now, as serializing an object referencing (for example) a complex object graph pretty much gives you the entire DB in serialized string format :-)

I'm fine with breaking BC here though: we had people serializing proxies since ages, and we always told them not to do it; time to stop that.

[beberlei]

Not implementing Proxy interface is a bit harsh, many third party code check for this, for example Symfony.

[Ocramius]

@beberlei I plan to implement Proxy, just not sure how (yet)

[Ocramius]

Just hooking in @alcaeus' suggestion for partial hydration of proxies: https://gist.github.com/alcaeus/fcc12672d7e46faa23e6c35516802e7b#file-proxy-php-L22

[Ocramius]

This patch needs to be re-applied on top of develop

[nsaleh]

Hey,

is there any update on this? Would be a great improvement in some contexts.

regards,
Nabil

[Ocramius]

@nsaleh I didn't work further on this, but I will do that as soon as I can. If you want to give it a shot yourself, be my guest!

[nsaleh]

Thanks for your quick answer and your work!, i might give it a try, but yeah time is a problem here too. If it would help and be appropriate to do maybe I could get some sponsoring, let me know.

As I understand this feature will be shipped with 3.0 is "done when it's done" right?

[Ocramius]

Shoot me an email and we'll see if it is feasible. I'm currently flooded with other work (good for the pockets, not for the time), which is why OSS currently falls behind

…

On 8 Aug 2017 10:55, "Nabil Saleh" ***@***.***> wrote: Thanks for your quick answer and your work!, i might give it a try, but yeah time is a problem here too. If it would help and be appropriate to do maybe I could get some sponsoring, let me know. As I understand this feature will be shipped with 3.0 is "done when it's done" right? — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#1241 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAJakIx2sTY28IXS5u2E2PvkjufqWRSCks5sWCJpgaJpZM4DQhEA> .

[Ocramius]

This was moved to #6719