global_namespace_import ignores imports for symbols that exist in code but ain't detected

[javaDeveloperKid]

Bug report

Description

global_namespace_import ignores classes (and constants?) that exist in code but ain't detected by php-cs-fixer. My case relates to PHPStan array shapes (see example below) but the problem is probably general.
I assume php-cs-fixer can understand array shapes because the rule no_unused_imports does not remove the use statement.

Expected behaviour
Either analyse PHPStan array shapes or at least report (non zero exit code?) that a use statement for global class/constant exists but it was not found (by php cs fixer) inside the code.

Runtime version

3.51.0

Used command

fix

Configuration file

'global_namespace_import' => ['import_classes' => false, 'import_constants' => false, 'import_functions' => false],
'no_unused_imports' => true,

Code snippet that reproduces the problem

<?php

namespace N;

use DateTimeImmutable;

class Foo
{
    /**
     * @param array{
     *     createdAt: DateTimeImmutable,
     * } $param1
     */
    function f(array $param1)
    {
    }
}

Expected

<?php

namespace N;

-use DateTimeImmutable;

class Foo
{
    /**
     * @param array{
-     *     createdAt: DateTimeImmutable,
+     *     createdAt: \DateTimeImmutable,
     * } $param1
     */
    function f(array $param1)
    {
    }
}

[Wirone]

@javaDeveloperKid please provide expected output, because from the report I just can't tell what's wrong. Used ./php-cs-fixer check --verbose --diff reproducer.php --rules='{"global_namespace_import":{"import_classes":false,"import_constants":false,"import_functions":false},"no_unused_imports":true}' for your snippet and got:

Should it be modified to createdAt: \DateTimeImmutable? 🤷‍♂️

Expected behaviour
Either analyse PHPStan array shapes or at least report (non zero exit code?) that a use statement for global class/constant exists but it was not found (by php cs fixer) inside the code.

Fixer exits with non-zero code only if dry-run is used and fixable violation is found. The core concept of this tool is to report only what can be fixed, it does not report any kind of "warnings".

[mvorisek]

I think this is duplicate of #7619

[Wirone]

Sort of, since it's related to different fixer 🙂.

[javaDeveloperKid]

Updated issue with expected output.

[mvorisek]

So it is related with #7674, FQCN fixer is quite complex as it has to parse a lot of different places, global_namespace_import should really reuse that logic.