Better change the PHPDoc to a fully qualified name than add an (unused) use statement.

@dunglas nope. Our coding standards are to use the short name in the phpdoc

👍

note to maintainers: this should go in 2.3

👍

Thank you @Miliooo.

@stof for my curiosity, do you have a link or an explanation of why to do that? Thanks.

By the way, the PSR-5 draft (PHPDoc) recommends to always use the Fully Qualified Class Name in DocBlock: https://github.com/php-fig/fig-standards/pull/169/files#diff-9a10a11696dc38209c125ea9c57a565fR1891

@dunglas the section you link is not telling that. It is saying that tools processing the phpdoc should resolve relative names to FQCN as soon as possible to make the processing easier (i.e. not having to remember the context). There is no recommendation at all about the way to write the class names in the phpdoc here

@stof yes I've misunderstood the statement. But why add a use statement just for the PHPDoc? Importing unused classes cannot have a bad impact on performances (I don't know, I just ask)?

@dunglas it makes PHPdocs a lot easier to read (especially when having namespaces as long as Symfony has).

@dunglas it cannot impact the performance if you use opcache. Use statements are resolved early in the compilation of the code, and this is done only once in prod when you can the opcodes

@stof thanks for the explanation :)

@stof are you sure this needed to be merged in the 2.3 branch.
See 0ab9c8d

I'm no expert in this but it seems this class was only added in 2.4.

@dunglas that is wrong. Classes are not autoloaded before they are actually used.
The use statement is nothing more than a hint for the following code so that the interpreter knows where to find your (short) class names.

See:
http://3v4l.org/rkp79

@apfelbox thanks for the link.

@apfelbox I'm not talking about the autoloading, but about the resolution of the name. Names are resolved even if the class does not exist.
Btw, this is exactly why a Java-like syntax use Foo\* cannot be implemented in PHP, as it would require knowing which classes exist in the Foo namespace.

@stof I know that, but @dunglas was talking about the autoloading. 😉

a.k.a. there is no performance impact, because there is no autoloading (exactly due to the reason you described: in PHP the direction is reversed: local class name -> FQN -> autoload instead of Java's import (= autoload) -> FQN -> local class name).

@apfelbox I've not talked about autoloading. Just about performances (and according to @stof there is a non-significative performance impact when using an opcode cache, and a more significative one when not using a cache).

Ah, okay, sorry - misunderstood.

But since the only thing affecting runtime performance here would be autoloading (and it doesn't happen), we can conclude that there is no impact. Right?