[RFC] Spread Operator in Array Expression View as Markdown

Hi, I worked together with Chris Wright, author of the other RFC covering
the same thing.

Thanks for pointing it out and also thanks CMB for posting the link.
I tried to search for this topic but didn't find anything, as we have too many names for this feature.

The most important thing is that we don't want it work like array_merge();
in other words, it must not cover both lists and maps at the same time.

Since PHP array is a mixture of list (traditional array in most of other languages) and map (a k=>v data structure),
any new syntax related to PHP array most cover both use cases.
... means unpacking, I think we are all good with this semantic, the disagreement is about the behavior of key conflict.

Let's use your example to start my explanation:

['a' => 1, 'b'=>2, ...['a' => 3]]

Since we are create a new syntax with the existing array definition syntax, let's try to do the unpack manually.

['a' => 1, 'b' =>2, 'a' => 3]

Although this not a good practice to write such codes, it's still valid in current version without any warning or even notice.
The answer for above code is

['a' => 3, 'b' => 2]

It looks weird at first glance, but IMO the more important thing is this is how PHP works now.
I really don't think it's a good idea to introduce a new way how two arrays will get merged.
It will confuse developers.

I also notice there is an interesting question in the discussion of your RFC.

So, it's just an operator version of array_merge(), with no advantages
other than brevity? I'm not sure why it's any more or less applicable to
constant arrays vs variables than + or array_merge() are.

Yes. The spread operator follows the semantics as array_merge, but besides brevity and better performance
(no more overhead of function call, and constant array can be compiled at compiled time),
this RFC also brings the support for objects implementing Traversables.

Best regards,
CHU Zhaowei