Skipping paths doesn't seem to work

[ArtemGoutsoul]

skip-path option does not seem to work (or I'm doing something wrong)

I have a folder structure like this (historical reasons):

• lib
  • vendor
  • other
  • composer.json
• app
  • reports
  • etc

I'd like to check lib and app folders and skip lib/vendor and app/reports

I tried tried:

php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path=lib/vendor,app/reports lib app
php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path=lib/vendor lib
php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path=app/reports app

The tool still scans the folders specified in skip-path.

Am I doing something wrong?

Thank you!

[TomasVotruba]

From top of my head, I think the option has to be used multiple times.

[ArtemGoutsoul]

thanks! I did try to use it just once, but didn't seem to work either

the reason why I notice is that some old libs in vendor folder have php templates that break the parsing:

php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path=lib/vendor lib

1. Finding used classes
=======================

  1156/24523 [▓░░░░░░░░░░░░░░░░░░░░░░░░░░░]   4%
In UseImportsResolver.php line 45:
                                                                                                                                                                    
  Could not parse file "/project/lib/vendor/coffeescript/coffeescript/vendor/ParserGenerator/PHP/Lempar.php": Syntax error, unexpected  
   '%' on line 97    

...

(note: I'm in /project folder when running the command)

[TomasVotruba]

Try --key value --key value, as our console is based symfony/console: https://symfony.com/doc/current/console/input.html#using-command-options

[ArtemGoutsoul]

I tried

php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path lib/vendor lib
php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path "lib/vendor" lib

same result as above

also tried changing the order:

php -d memory_limit=1024M lib/vendor/bin/class-leak check lib --skip-path lib/vendor
php -d memory_limit=1024M lib/vendor/bin/class-leak check lib --skip-path "lib/vendor"

same result (trying to parse files in lib/vendor)

[Lc5]

@ArtemGoutsoul You can try: php -d memory_limit=1024M lib/vendor/bin/class-leak check --skip-path=vendor --skip-path=lib

[FeBe95]

I've found the reason for this unexpected behavior: The symfony/finder package simply does not support this type of relative directory exclusion.

Simple Example

Let's examine the following command:

vendor/bin/class-leak check --skip-path=lib/vendor -- lib

The check command now does the following:

1. As mentioned above, class-leak uses the symfony/finder package for it's file/directory scanning
2. The passed path argument (lib) gets set as the $rootPath property of the scanner
   see: symfony/finder/Iterator/RecursiveDirectoryIterator.php#L45
3. From now on all relative file names of the iterator will be relative to this $rootPath
4. Any exclusions will be matched against these relative file names, that do not contain the root path itself
   see: symfony/finder/Iterator/ExcludeDirectoryFilterIterator.php#L80

Let's imagine a directory structure like this:

MyProjectRoot # "check" command is being executed from this location
└─ lib
   └─ vendor
      └─ someDir
         └─ someFile.php

When the iterator check this file, all it sees is this:

• Root path: lib
• Relative file path: vendor/someDir/someFile.php
• Exclusion pattern: lib/vendor (plus some regex chars)
• Result: ❌ The iterator does not exclude this file

Workaround

To successfully exclude files in lib/vendor the command needs to be this instead:

vendor/bin/class-leak check --skip-path=vendor -- lib

Complex Example

Now, let's pass two paths to each argument/option at the same time:

vendor/bin/class-leak check --skip-path=vendor --skip-path=tests -- app lib

What will happen now?
The directory iterator will exclude all files in these directories:

• MyProjectRoot/app/tests/*
• MyProjectRoot/app/vendor/*
• MyProjectRoot/lib/tests/*
• MyProjectRoot/lib/vendor/*

This is not what we want, I think.

Note

There are a lot of ongoing discussions on how to handle the ->exclude() method over at the symfony/symfony repo:

• There is no way to exclude path only from root directory symfony/symfony#28158
• Finder excluded does NOT exclude relative directories symfony/symfony#47431
• [Finder] Exclude relative to content root if prefixed / symfony/symfony#54752
• [Finder] Exclude relative to content root if prefixed / symfony/symfony#54754

Solution

The scanner logic needs to be rewritten to allow for "relative" path exclusions. The issues for the Symfony Finder package date back to 2018. I don't know if any of the two open pull requests will be merged soon.

In order to support the "multiple paths with multiple skip-paths" case we need to implement a custom filter solution. To be specific, we could change this logic:

class-leak/src/Commands/CheckCommand.php

Lines 118 to 119 in d4ad8f0

	// but we only want to check the files that are not in the skipped paths
	$phpFilePaths = $this->phpFilesFinder->findPhpFiles($paths, $fileExtensions, $pathsToSkip);

Instead of running the Finder a second time, just iterate through the existing path list and remove matching skip-path directories from the list.

@TomasVotruba What do you think?

[FeBe95]

Here's a rough implementation that works for this specific use case. I don't know if this breaks any other stuff though.

// but we only want to check the files that are not in the skipped paths
$phpFilePaths = $allFilePaths;
$basePath = rtrim(str_replace('\\', '/', getcwd()), '/');

foreach ($pathsToSkip as $pathToSkip) {
    $pathToSkip = str_replace('\\', '/', $pathToSkip) . '/';

    $phpFilePaths = array_filter($phpFilePaths, function ($filePath) use ($basePath, $pathToSkip) {
        $path = str_replace('\\', '/', $filePath);
        $excludePath = $basePath . '/'. $pathToSkip;

        return !str_starts_with($path, $excludePath);
    });
}