Allow to pass a thunk as a message factory to assertions

[simPod]

It is useful in order to delay generating a message.

At the moment the assertion does not pass the message is generated. Otherwise the thunk is not executed.

[simPod]

I did not implement it to all assertions. If RFC is accepted, I'll add it everywhere

[shadowhand]

Actually, looking at this again, I wonder if using Stringable|string would be more appropriate, thus dropping the need for a closure.

[simPod]

How would you expect it to work?

my proposal is

Assert::string(1, fn() => heavy compute here)

With Stringable it has to be more verbose

Assert::string(1, new class() { public function __toString() { ... } })

[shadowhand]

@simPod yes, more verbose. And also more type safe, given that callable(): string is not enforceable (without static analysis) but Stringable|string is. 🤔

[shadowhand]

In any case, I am not merging new features right now, but this is high priority for the next feature release.

[simPod]

Same as array<object|string> is not enforceable in isInstanceOfAny, that's not valid argument IMO.

I'm not asking for merge now

[zerkms]

What would be the real life use case scenario for this?

[simPod]

Sometimes my error messages are

1. heavy to compute
2. touch object properties that are not initialised and are not needed to be initialised except for the error message

so I'd rather skip that for happy path.

[simPod]

@BackEndTea Hi, is there a will to add lazy messages?

[shadowhand]

Interesting idea, and certainly something I would consider for the next feature release.