Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:111499
To: Claude Pache <claude.pache@gmail.com>
Cc: "internals@lists.php.net" <internals@lists.php.net>
In-Reply-To: <A282700A-A2E5-4757-B611-B7E41B728B7B@gmail.com>
References: <641541597256204@mail.yandex.com> <A282700A-A2E5-4757-B611-B7E41B728B7B@gmail.com>
MIME-Version: 1.0
Date: Wed, 12 Aug 2020 17:05:11 -0300
Message-ID: <1252211597259017@mail.yandex.com>
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset=utf-8
Subject: [PHP-DEV] Nested functions
From: luis@softsatsistemas.com.br (Luis - SoftSAT Sistemas)

>>  Le 12 août 2020 à 20:19, Luis - SoftSAT Sistemas <luis@softsatsistemas.com.br> a écrit :
>>
>>  In the effort to make PHP a better language i want to ask why not "fix" nested functions as suggested in (https://wiki.php.net/rfc/true-nested-function-support) or block their usage?
>>  The current behavior is a lot unexpected.
>>
>>  I am new here, sorry if it was already discussed, i can't find a good way to search the mailing history.
>
> Hi,
>
> Defining a global function inside another function is not useless. Here is a dummy example:
>
> ```php
> // a.php
> spl_autoload_register(function ($classname) {
>     if ($classname === 'Foo\\B')
>         require 'foo/b.php';
> })
>
> // foo/b.php
> namespace Foo;
>
> class B {
>     // ...
> }
>
> function some_helper(....$args) {
>     // ...
> }
> ```
>
> The global function `Foo\some_helper` is defined inside the function used as autoloader.
>
> Also, from the RFC:
>
>>  They exist in the global namespace.
>
> I think you meant “They exist in the global scope”? because they are defined in the namespace corresponding to the eventual `namespace` declaration of the file.
>
> —Claude

You are right Claude, i wasn't thinking about this use case.

But outside of this use case more problems arise.
If `foo/b.php`  have some global variable and some function that use it the function breaks because the global variable will be "included" into current scope, while the functions into namespace scope.
```php
//a.php
function foo() {
    require 'b.php';
    //var_dump($something);
}

foo();

var_dump(bar());

//b.php
$something = 10;

function bar() {
    global $something;
    return $something;
}
```
I expected the `var_dump(bar());` output to be `int (10)`, but it was `NULL`, while if i do `var_dump($something);` from `foo()` it outputs `int(10)` as "normally expected".
Given that "bar()" are at namespace scope, this can be "fixed" changing `$something = 10;` to `$GLOBALS['something'] = 10;`, but that doesn't change the fact that this behavior are odd and not intuitive.

PS: I known that global variables are a bad thing, it was just an example.

Anyway now i see that changing this will break the auto loader system.

About the RFC you are right, they are wrong about the scope, but the RFC aren't mine, i just found it while searching about the subject.

--L. Henrique