"regular expression is too large" when handling large arrays

[kingkero]

Bug report

Currently introducing PHPStan to a larger Laminas project (already using slam/phpstan-laminas-framework), when generating the baseline one entry is:

    -
      message: "#^Cannot access offset 'grpc' on 'https\\://…'\\|'https\\://…'\\|\\(array\\<int\\|\\(literal\\-string&non\\-falsy\\-string\\), 0\\|0\\.25\\|0\\.66\\|1\\|3\\|5\\|10\\|14\\|15\\|20\\|29\\|50\\|60\\|73\\|74\\|75\\|76\\|79\\|80\\|99\\|100\\|123\\|124\\|260\\|300\\|500\\|600\\|660\\|1000\\|2000\\|3600\\|4000\\|9002\\|10000\\|11211\\|24575\\|''\\|'/'\\|'/api/v1/'\\|'/api/v2/'\\|'/api/v3/'\\|'/backup/…'\\|'/collect'\\|'/overload_fp\\.html'\\|'/var/www/bigassets…'\\|'/var/www/…'\\|'/var/www/…'\\|'/var/www/|false\\.|HERE THE REGEX CONTINUES FOR VERY LONG$#"
      count: 3 
      path: ../../../module/src/MyService.php

I already shortened the line. The original line in the baseline is 60733 characters long.

Running phpstan analyse with that baseline results in

In Strings.php line 447:
                                                                                                                                                                                                                                                                           
  [_PHPStan_156ee64ba\Nette\Utils\RegexpException]                                                                                                                                                                                                                         
  Compilation failed: regular expression is too large at offset 53891 in pattern: #^Cannot access offset 'grpc' on 'https\://…'\|'https\://…'\|\(array<int\|\(literal\-string...HERE THE LONG REGEX IS LISTED FROM BASELINE

Exception trace:
  at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/nette/utils/src/Utils/Strings.php:447
 _PHPStan_156ee64ba\Nette\Utils\Strings::_PHPStan_156ee64ba\Nette\Utils\{closure}() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/nette/utils/src/Utils/Callback.php:63
 _PHPStan_156ee64ba\Nette\Utils\Callback::_PHPStan_156ee64ba\Nette\Utils\{closure}() at n/a:n/a
 preg_match() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/nette/utils/src/Utils/Callback.php:70
 _PHPStan_156ee64ba\Nette\Utils\Callback::invokeSafe() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/nette/utils/src/Utils/Strings.php:448
 _PHPStan_156ee64ba\Nette\Utils\Strings::pcre() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/nette/utils/src/Utils/Strings.php:410
 _PHPStan_156ee64ba\Nette\Utils\Strings::match() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Analyser/IgnoredError.php:47
 PHPStan\Analyser\IgnoredError::shouldIgnore() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Analyser/IgnoredErrorHelperResult.php:83
 PHPStan\Analyser\IgnoredErrorHelperResult->PHPStan\Analyser\{closure}() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Analyser/IgnoredErrorHelperResult.php:139
 PHPStan\Analyser\IgnoredErrorHelperResult->PHPStan\Analyser\{closure}() at n/a:n/a
 array_filter() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Analyser/IgnoredErrorHelperResult.php:168
 PHPStan\Analyser\IgnoredErrorHelperResult->process() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Command/AnalyseApplication.php:120
 PHPStan\Command\AnalyseApplication->analyse() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/src/Command/AnalyseCommand.php:200
 PHPStan\Command\AnalyseCommand->execute() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/symfony/console/Command/Command.php:259
 _PHPStan_156ee64ba\Symfony\Component\Console\Command\Command->run() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/symfony/console/Application.php:870
 _PHPStan_156ee64ba\Symfony\Component\Console\Application->doRunCommand() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/symfony/console/Application.php:261
 _PHPStan_156ee64ba\Symfony\Component\Console\Application->doRun() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/vendor/symfony/console/Application.php:157
 _PHPStan_156ee64ba\Symfony\Component\Console\Application->run() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/bin/phpstan:124
 _PHPStan_156ee64ba\{closure}() at phar:///var/www/vendor/phpstan/phpstan/phpstan.phar/bin/phpstan:125
 require() at /var/www/vendor/phpstan/phpstan/phpstan:8
 include() at /var/www/vendor/bin/phpstan:119

The line in code that triggers this error is:

$config = $this->serviceManager->get('config');

$hostname = $config['...']['grpc']['hostname'];
$grpcOptions = $config['...']['grpc']['options'];
$channel = $config['...']['grpc']['channel'];

This issue only happens when using the baseline. If I have an empty baseline, generate it new - it runs perfectly. But after generation trying to re-run I get the above posted issue.

I tried reproducing the issue in the playground, but the playground crashes. The code used can be found here (could not post it in the issue).

Code snippet that reproduces the problem

No response

Expected output

PHPStan doesn't crash parsing baselines it previously generated.

Did PHPStan help you today? Did it make you happy in any way?

No response

[mergeable]

This bug report is missing a link to reproduction at phpstan.org/try.

It will most likely be closed after manual review.

[ondrejmirtes]

Oh my god, what in the hell is that code!? Doesn't look like a maintainable piece of code. My smartphone almost crashed trying to load the page.

I'm inclined to say that this kind of bending and hacking is not supported 😊

[kingkero]

Honestly I'm not sure if that example fits. I'm not sure who generates that long regex, probably some magic from the Laminas connector that tries to set all config keys it finds or something. It is not something part of our codebase.

The issue is that PHPStan generates a baseline entry it can't parse.

[ondrejmirtes]

I tried to make the problem less worse: phpstan/phpstan-src@039a3dc

Please try composer require --dev phpstan/phpstan:1.10.x-dev and run PHPStan, generate the baseline and use the baseline. Thanks.

[ondrejmirtes]

I'll reopen if you find out it still doesn't work.

[kingkero]

@ondrejmirtes Sorry, but cannot confirm the fix. Not sure if its a version mis-match (I am behind a packagist proxy), but even cloning phpstan-src and compiling the phar, I still saw the issue.

Could create a reproduction repo for this: https://github.com/kingkero/phpstan-error-too-long/, info in README.

Basically this (https://github.com/kingkero/phpstan-error-too-long/blob/main/module/Application/config/module.config.php#L8-L11) simulates the long config that we have in the prod setup.

[ondrejmirtes]

@kingkero Hi, I verified that the fix works. But you misunderstood the nature of the fix. You need to delete your current baseline that's commited here https://github.com/kingkero/phpstan-error-too-long/blob/main/phpstan.baseline.neon and generate it again with the newer PHPStan dev version. It's going to contain a shorter error message and the regex machine is not going to have a problem with that.

You can install that version with composer require --dev phpstan/phpstan:1.10.x-dev.

[kingkero]

You're right, I'm sorry for the extra noise. Can confirm it works as described.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.