Treat namespaced names as single token, allow using reserved class names

[nikic]

RFC: https://wiki.php.net/rfc/namespaced_names_as_token

Instead of treating namespaced names a sequence of T_STRING and T_NS_SEPARATOR tokens, treat them as a single token. In particular, this allows the use of reserved keywords inside namespaced names. At the same time allow using reserved keywords as class, function and constant names -- these cannot be directly used, but they can be used inside a namespaced name (e.g. fully qualified).

This is BC breaking, because whitespace between namespace separators is no longer supported, so Foo \ Bar is no longer considered a legal namespaced name. I consider this a feature, because this avoids bugs that I've seen a couple of times, like:

// Previously this generated an undefined constant diagnostic,
// which is pretty hard to understand if your not familiar with PHP's
// obscure "namespace-relative name" feature.
namespace \Foo;

$x = Foo // <- semicolon missing here
\strlen($bar);

// Interpreted as
$x = Foo\strlen($bar);

[nikic]

Regressions found in top 2k composer packages due to forbidden whitespace in namespaced names:

sylius/sylius/src/Sylius/Bundle/ApiBundle/ApiPlatform/Metadata/Property/Factory/ExtractorPropertyMetadataFactory.php:109
    \ RuntimeException
api-platform/core/src/Metadata/Extractor/AbstractExtractor.php:121
    \ RuntimeException
mck89/peast/lib/Peast/Syntax/Node/JSX/JSXFragment.php:13
    Peast \Syntax\Node\Expression
mck89/peast/lib/Peast/Syntax/Node/JSX/JSXOpeningElement.php:13
    Peast \Syntax\Node\Expression
mck89/peast/lib/Peast/Syntax/Node/JSX/JSXElement.php:13
    Peast \Syntax\Node\Expression

[marcioAlmada]

Hello!

I remember this was the only obstacle that impaired a second version of the context sensitive language RFC, but was considered a bigger BC break at the time. After that change and later parser changes to allow classes, interfaces, functions, traits and constants to have any name I believe only a single edge case happened:

https://3v4l.org/vUVjn

namespace Namespace; // still not possible

namespace/Foo:class; // due to this alternative namespace resolution construct
Foo:class; // which is equivalent to this anyway xD

But that wouldn't be a big deal, probably.

[marcioAlmada]

BTW https://3v4l.org/bOvQJ some constructs re-declared as global functions, could lead to some problems:

// troll.php
function echo(){}
function die(){}
function exit(){}

//

echo('bye');
die('now!');
exit('please?!')

... but also could help making untestable code more testable xD

[marcioAlmada]

Maybe an error like "Can not redeclare intrinsic PHP construct echo..." makes more sense here? Same for exit, die and print ?

[nikic]

@marcioAlmada I'm intentionally allowing all of these, without special cases. All symbols can be used, they just need to referenced though a namespaced name, such as FooBar\list().

Or is the suggestion here specifically to forbid the function-like language constructs in the global scope only?

[marcioAlmada]

@marcioAlmada I'm intentionally allowing all of these, without special cases. All symbols can be used, they just need to referenced though a namespaced name, such as FooBar\list().

Or is the suggestion here specifically to forbid the function-like language constructs in the global scope only?

Definitely just function-like language stuff like die.

Or perhaps if declaration is allowed at least don't override PHP die() "call" unless the userland function is explicitly called as \die();

[nikic]

Closing in favor of #5827.