[RFC] Allow Readonly Property Defaults

[NickSdot]

RFC: https://wiki.php.net/rfc/readonly_property_defaults
Discussion thread: https://news-web.php.net/php.internals/131760
Voting thread: tbd

[TimWolla]

It is legal for properties to appear twice in serialization payload?! This probably should be deprecated and killed.

[NickSdot]

Same answer here, the test is also just copied from 12 lines above. Intention was to confirm that having a default value makes no difference. Do you want me to change something? How?

[TimWolla]

This was an unrelated remark about serialization in general. Often one notices odd behavior about PHP while reviewing PRs.

No action needed from your side.

[NickSdot]

Ref: #22599

[TimWolla]

Looking at this once more after reading the RFC (and providing feedback on-list): I think this test is broken and effectively testing an implementation detail that is not usually observable.

The duplicate property is actually written twice (and readonly ignored) when no __unserialize() hook exists, but when an __unserialize() hook exists, only the latter version for the same key will be seen by the hook.

And similarly for a readonly property with a default value (i.e. the new feature), without __unserialize() the default will be overwritten, but __unserialize() will not be able to set it, because it is beholden to the regular readonly rules.

When you add:

    public function __unserialize(array $data): void
    {
        foreach ($data as $key => $val) {
                $this->{$key} = $val;
        }
    }

then you'll get:

Error: Cannot modify readonly property TestDefault::$prop in