[5.4] Don't reverse the pushed data

[jbrooksuk]

Closes #14876

[taylorotwell]

I feel like we must have been doing this for a reason? 😄

[jbrooksuk]

@taylorotwell I originally thought that, but the change that introduced the reversal didn't explicitly state why.

[taylorotwell]

Would like to get to the bottom of that if possible.

[taylorotwell]

@mark86092 can you let us know why you introduced an array_reverse on the push?

[jbrooksuk]

Pinging @mark86092?

[jbrooksuk]

Oh, you beat me.

[mark86092]

Wait me a moment, I'll reply soon

[mark86092]

The reason goes to several statements:

1. Rendering Ordering:

The first is child --> then parent --> then the grand parent blade template

with renderCount increases to identify

So, the PR #12808 's tests use the $factory->incrementRender(); in L285, to mimic the @push in parent.

2. The Push Content Ordering:

I would expect the @push content in parent should appear before child, and the @push content in grand parent should appear before parent blade template

Reason from 1) and 2)

So, I think when render the $this->pushContentsStack[$section], it is better to array_reverse it.

Making the parent's push content (i.e. $this->pushContentsStack[$section][1]) appears before the child's push content (i.e. $this->pushContentsStack[$section][0])

[mark86092]

During the investigation of the reason, I found that in BladeCompiler.php, compileExtends and compileInclude share almost the same render behaviors. Could it be the reason why array_reverse or not array_reverse ?

[taylorotwell]

@jbrooksuk any thoughts on that? 😄

[taylorotwell]

I tend to agree with removing this. It becomes more apparent the behavior is wrong when you use something like the new 5.4 components and slots... your pushes are in reverse order when you dump them in that situation.

[taylorotwell]

I think there are situations where the array_reverse behavior makes sense but perhaps it needs to be thought out more. To me push should render its contents in the order they were pushed. Otherwise its not a push its just a prepend.

[jbrooksuk]

Sorry, @taylorotwell I missed your message. But yeah, slots would make the reverse weirder.

[rico-ocepek]

From my point of view the 5.3 order was better (maybe not correct but better).
Just updated to 5.4 and all my javascript inclusions broke due to the child's javascript being rendered before the parent's.
It's probably the correct behavior for something called 'stack' but I really wish for something queue-like now.

For the moment i fixed my error using prioritized yields for the scripts that need to be loaded first but this doesn't feel nice at all.

Any ideas how to work arround this?